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
579
specified. They are returned in a partial order by compression
580
parent, so that the deltas can be applied as the data stream is
581
inserted; however note that compression parents will not be sent
582
unless they were specifically requested, as the client may already
585
:param required_versions: The exact set of versions to be extracted.
586
Unlike some other knit methods, this is not used to generate a
587
transitive closure, rather it is used precisely as given.
589
:returns: format_signature, list of (version, options, length, parents),
592
required_version_set = frozenset(required_versions)
594
# list of revisions that can just be sent without waiting for their
597
# map from revision to the children based on it
599
# first, read all relevant index data, enough to sort into the right
601
for version_id in required_versions:
602
options = self._index.get_options(version_id)
603
parents = self._index.get_parents_with_ghosts(version_id)
604
index_memo = self._index.get_position(version_id)
605
version_index[version_id] = (index_memo, options, parents)
606
if ('line-delta' in options
607
and parents[0] in required_version_set):
608
# must wait until the parent has been sent
609
deferred.setdefault(parents[0], []). \
612
# either a fulltext, or a delta whose parent the client did
613
# not ask for and presumably already has
614
ready_to_send.append(version_id)
615
# build a list of results to return, plus instructions for data to
617
copy_queue_records = []
618
temp_version_list = []
620
# XXX: pushing and popping lists may be a bit inefficient
621
version_id = ready_to_send.pop(0)
622
(index_memo, options, parents) = version_index[version_id]
623
copy_queue_records.append((version_id, index_memo))
624
none, data_pos, data_size = index_memo
625
temp_version_list.append((version_id, options, data_size,
627
if version_id in deferred:
628
# now we can send all the children of this revision - we could
629
# put them in anywhere, but we hope that sending them soon
630
# after the fulltext will give good locality in the receiver
631
ready_to_send[:0] = deferred.pop(version_id)
632
assert len(deferred) == 0, \
633
"Still have compressed child versions waiting to be sent"
634
# XXX: The stream format is such that we cannot stream it - we have to
635
# know the length of all the data a-priori.
637
result_version_list = []
638
for (version_id, raw_data), \
639
(version_id2, options, _, parents) in \
640
izip(self._data.read_records_iter_raw(copy_queue_records),
642
assert version_id == version_id2, \
643
'logic error, inconsistent results'
644
raw_datum.append(raw_data)
645
result_version_list.append(
646
(version_id, options, len(raw_data), parents))
647
# provide a callback to get data incrementally.
648
pseudo_file = StringIO(''.join(raw_datum))
651
return pseudo_file.read()
653
return pseudo_file.read(length)
654
return (self.get_format_signature(), result_version_list, read)
656
def _extract_blocks(self, version_id, source, target):
657
if self._index.get_method(version_id) != 'line-delta':
659
parent, sha1, noeol, delta = self.get_delta(version_id)
660
return KnitContent.get_line_delta_blocks(delta, source, target)
662
def get_delta(self, version_id):
663
"""Get a delta for constructing version from some other version."""
664
self.check_not_reserved_id(version_id)
665
parents = self.get_parents(version_id)
670
index_memo = self._index.get_position(version_id)
671
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
672
noeol = 'no-eol' in self._index.get_options(version_id)
673
if 'fulltext' == self._index.get_method(version_id):
674
new_content = self.factory.parse_fulltext(data, version_id)
675
if parent is not None:
676
reference_content = self._get_content(parent)
677
old_texts = reference_content.text()
680
new_texts = new_content.text()
681
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
683
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
685
delta = self.factory.parse_line_delta(data, version_id)
686
return parent, sha1, noeol, delta
688
def get_format_signature(self):
689
"""See VersionedFile.get_format_signature()."""
690
if self.factory.annotated:
691
annotated_part = "annotated"
693
annotated_part = "plain"
694
return "knit-%s" % (annotated_part,)
696
def get_graph_with_ghosts(self):
697
"""See VersionedFile.get_graph_with_ghosts()."""
698
graph_items = self._index.get_graph()
699
return dict(graph_items)
701
def get_sha1(self, version_id):
702
return self.get_sha1s([version_id])[0]
704
def get_sha1s(self, version_ids):
705
"""See VersionedFile.get_sha1()."""
706
record_map = self._get_record_map(version_ids)
707
# record entry 2 is the 'digest'.
708
return [record_map[v][2] for v in version_ids]
712
"""See VersionedFile.get_suffixes()."""
713
return [DATA_SUFFIX, INDEX_SUFFIX]
715
def has_ghost(self, version_id):
716
"""True if there is a ghost reference in the file to version_id."""
718
if self.has_version(version_id):
720
# optimisable if needed by memoising the _ghosts set.
721
items = self._index.get_graph()
722
for node, parents in items:
723
for parent in parents:
724
if parent not in self._index._cache:
725
if parent == version_id:
729
def insert_data_stream(self, (format, data_list, reader_callable)):
730
"""Insert knit records from a data stream into this knit.
732
If a version in the stream is already present in this knit, it will not
733
be inserted a second time. It will be checked for consistency with the
734
stored version however, and may cause a KnitCorrupt error to be raised
735
if the data in the stream disagrees with the already stored data.
737
:seealso: get_data_stream
739
if format != self.get_format_signature():
740
trace.mutter('incompatible format signature inserting to %r', self)
741
raise KnitDataStreamIncompatible(
742
format, self.get_format_signature())
744
for version_id, options, length, parents in data_list:
745
if self.has_version(version_id):
746
# First check: the list of parents.
747
my_parents = self.get_parents_with_ghosts(version_id)
748
if my_parents != parents:
749
# XXX: KnitCorrupt is not quite the right exception here.
752
'parents list %r from data stream does not match '
753
'already recorded parents %r for %s'
754
% (parents, my_parents, version_id))
756
# Also check the SHA-1 of the fulltext this content will
758
raw_data = reader_callable(length)
759
my_fulltext_sha1 = self.get_sha1(version_id)
760
df, rec = self._data._parse_record_header(version_id, raw_data)
761
stream_fulltext_sha1 = rec[3]
762
if my_fulltext_sha1 != stream_fulltext_sha1:
763
# Actually, we don't know if it's this knit that's corrupt,
764
# or the data stream we're trying to insert.
766
self.filename, 'sha-1 does not match %s' % version_id)
768
if 'line-delta' in options:
769
# Make sure that this knit record is actually useful: a
770
# line-delta is no use unless we have its parent.
771
# Fetching from a broken repository with this problem
772
# shouldn't break the target repository.
773
if not self._index.has_version(parents[0]):
776
'line-delta from stream references '
777
'missing parent %s' % parents[0])
778
self._add_raw_records(
779
[(version_id, options, parents, length)],
780
reader_callable(length))
783
"""See VersionedFile.versions."""
784
if 'evil' in debug.debug_flags:
785
trace.mutter_callsite(2, "versions scales with size of history")
786
return self._index.get_versions()
788
def has_version(self, version_id):
789
"""See VersionedFile.has_version."""
790
if 'evil' in debug.debug_flags:
791
trace.mutter_callsite(2, "has_version is a LBYL scenario")
792
return self._index.has_version(version_id)
794
__contains__ = has_version
796
def _merge_annotations(self, content, parents, parent_texts={},
797
delta=None, annotated=None,
798
left_matching_blocks=None):
799
"""Merge annotations for content. This is done by comparing
800
the annotations based on changed to the text.
802
if left_matching_blocks is not None:
803
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
807
for parent_id in parents:
808
merge_content = self._get_content(parent_id, parent_texts)
809
if (parent_id == parents[0] and delta_seq is not None):
812
seq = patiencediff.PatienceSequenceMatcher(
813
None, merge_content.text(), content.text())
814
for i, j, n in seq.get_matching_blocks():
817
# this appears to copy (origin, text) pairs across to the
818
# new content for any line that matches the last-checked
820
content._lines[j:j+n] = merge_content._lines[i:i+n]
822
if delta_seq is None:
823
reference_content = self._get_content(parents[0], parent_texts)
824
new_texts = content.text()
825
old_texts = reference_content.text()
826
delta_seq = patiencediff.PatienceSequenceMatcher(
827
None, old_texts, new_texts)
828
return self._make_line_delta(delta_seq, content)
830
def _make_line_delta(self, delta_seq, new_content):
831
"""Generate a line delta from delta_seq and new_content."""
833
for op in delta_seq.get_opcodes():
836
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
839
def _get_components_positions(self, version_ids):
840
"""Produce a map of position data for the components of versions.
842
This data is intended to be used for retrieving the knit records.
844
A dict of version_id to (method, data_pos, data_size, next) is
846
method is the way referenced data should be applied.
847
data_pos is the position of the data in the knit.
848
data_size is the size of the data in the knit.
849
next is the build-parent of the version, or None for fulltexts.
852
for version_id in version_ids:
855
while cursor is not None and cursor not in component_data:
856
method = self._index.get_method(cursor)
857
if method == 'fulltext':
860
next = self.get_parents_with_ghosts(cursor)[0]
861
index_memo = self._index.get_position(cursor)
862
component_data[cursor] = (method, index_memo, next)
864
return component_data
866
def _get_content(self, version_id, parent_texts={}):
867
"""Returns a content object that makes up the specified
869
cached_version = parent_texts.get(version_id, None)
870
if cached_version is not None:
871
if not self.has_version(version_id):
872
raise RevisionNotPresent(version_id, self.filename)
873
return cached_version
875
text_map, contents_map = self._get_content_maps([version_id])
876
return contents_map[version_id]
878
def _check_versions_present(self, version_ids):
879
"""Check that all specified versions are present."""
880
self._index.check_versions_present(version_ids)
882
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
883
nostore_sha, random_id, check_content):
884
"""See VersionedFile.add_lines_with_ghosts()."""
885
self._check_add(version_id, lines, random_id, check_content)
886
return self._add(version_id, lines, parents, self.delta,
887
parent_texts, None, nostore_sha, random_id)
889
def _add_lines(self, version_id, parents, lines, parent_texts,
890
left_matching_blocks, nostore_sha, random_id, check_content):
891
"""See VersionedFile.add_lines."""
892
self._check_add(version_id, lines, random_id, check_content)
893
self._check_versions_present(parents)
894
return self._add(version_id, lines[:], parents, self.delta,
895
parent_texts, left_matching_blocks, nostore_sha, random_id)
897
def _check_add(self, version_id, lines, random_id, check_content):
898
"""check that version_id and lines are safe to add."""
899
if contains_whitespace(version_id):
900
raise InvalidRevisionId(version_id, self.filename)
901
self.check_not_reserved_id(version_id)
902
# Technically this could be avoided if we are happy to allow duplicate
903
# id insertion when other things than bzr core insert texts, but it
904
# seems useful for folk using the knit api directly to have some safety
905
# blanket that we can disable.
906
if not random_id and self.has_version(version_id):
907
raise RevisionAlreadyPresent(version_id, self.filename)
909
self._check_lines_not_unicode(lines)
910
self._check_lines_are_lines(lines)
912
def _add(self, version_id, lines, parents, delta, parent_texts,
913
left_matching_blocks, nostore_sha, random_id):
914
"""Add a set of lines on top of version specified by parents.
916
If delta is true, compress the text as a line-delta against
919
Any versions not present will be converted into ghosts.
921
# first thing, if the content is something we don't need to store, find
923
line_bytes = ''.join(lines)
924
digest = sha_string(line_bytes)
925
if nostore_sha == digest:
926
raise errors.ExistingContent
929
if parent_texts is None:
931
for parent in parents:
932
if self.has_version(parent):
933
present_parents.append(parent)
935
# can only compress against the left most present parent.
937
(len(present_parents) == 0 or
938
present_parents[0] != parents[0])):
941
text_length = len(line_bytes)
944
if lines[-1][-1] != '\n':
945
# copy the contents of lines.
947
options.append('no-eol')
948
lines[-1] = lines[-1] + '\n'
952
# To speed the extract of texts the delta chain is limited
953
# to a fixed number of deltas. This should minimize both
954
# I/O and the time spend applying deltas.
955
delta = self._check_should_delta(present_parents)
957
assert isinstance(version_id, str)
958
content = self.factory.make(lines, version_id)
959
if delta or (self.factory.annotated and len(present_parents) > 0):
960
# Merge annotations from parent texts if needed.
961
delta_hunks = self._merge_annotations(content, present_parents,
962
parent_texts, delta, self.factory.annotated,
963
left_matching_blocks)
966
options.append('line-delta')
967
store_lines = self.factory.lower_line_delta(delta_hunks)
968
size, bytes = self._data._record_to_data(version_id, digest,
971
options.append('fulltext')
972
# isinstance is slower and we have no hierarchy.
973
if self.factory.__class__ == KnitPlainFactory:
974
# Use the already joined bytes saving iteration time in
976
size, bytes = self._data._record_to_data(version_id, digest,
979
# get mixed annotation + content and feed it into the
981
store_lines = self.factory.lower_fulltext(content)
982
size, bytes = self._data._record_to_data(version_id, digest,
985
access_memo = self._data.add_raw_records([size], bytes)[0]
986
self._index.add_versions(
987
((version_id, options, access_memo, parents),),
989
return digest, text_length, content
991
def check(self, progress_bar=None):
992
"""See VersionedFile.check()."""
994
def _clone_text(self, new_version_id, old_version_id, parents):
995
"""See VersionedFile.clone_text()."""
996
# FIXME RBC 20060228 make fast by only inserting an index with null
998
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1000
def get_lines(self, version_id):
1001
"""See VersionedFile.get_lines()."""
1002
return self.get_line_list([version_id])[0]
1004
def _get_record_map(self, version_ids):
1005
"""Produce a dictionary of knit records.
1007
The keys are version_ids, the values are tuples of (method, content,
1009
method is the way the content should be applied.
1010
content is a KnitContent object.
1011
digest is the SHA1 digest of this version id after all steps are done
1012
next is the build-parent of the version, i.e. the leftmost ancestor.
1013
If the method is fulltext, next will be None.
1015
position_map = self._get_components_positions(version_ids)
1016
# c = component_id, m = method, i_m = index_memo, n = next
1017
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1019
for component_id, content, digest in \
1020
self._data.read_records_iter(records):
1021
method, index_memo, next = position_map[component_id]
1022
record_map[component_id] = method, content, digest, next
1026
def get_text(self, version_id):
1027
"""See VersionedFile.get_text"""
1028
return self.get_texts([version_id])[0]
1030
def get_texts(self, version_ids):
1031
return [''.join(l) for l in self.get_line_list(version_ids)]
1033
def get_line_list(self, version_ids):
1034
"""Return the texts of listed versions as a list of strings."""
1035
for version_id in version_ids:
1036
self.check_not_reserved_id(version_id)
1037
text_map, content_map = self._get_content_maps(version_ids)
1038
return [text_map[v] for v in version_ids]
1040
_get_lf_split_line_list = get_line_list
1042
def _get_content_maps(self, version_ids):
1043
"""Produce maps of text and KnitContents
1045
:return: (text_map, content_map) where text_map contains the texts for
1046
the requested versions and content_map contains the KnitContents.
1047
Both dicts take version_ids as their keys.
1049
# FUTURE: This function could be improved for the 'extract many' case
1050
# by tracking each component and only doing the copy when the number of
1051
# children than need to apply delta's to it is > 1 or it is part of the
1053
version_ids = list(version_ids)
1054
multiple_versions = len(version_ids) != 1
1055
record_map = self._get_record_map(version_ids)
1060
for version_id in version_ids:
1063
while cursor is not None:
1064
method, data, digest, next = record_map[cursor]
1065
components.append((cursor, method, data, digest))
1066
if cursor in content_map:
1071
for component_id, method, data, digest in reversed(components):
1072
if component_id in content_map:
1073
content = content_map[component_id]
1075
if method == 'fulltext':
1076
assert content is None
1077
content = self.factory.parse_fulltext(data, version_id)
1078
elif method == 'line-delta':
1079
delta = self.factory.parse_line_delta(data, version_id)
1080
if multiple_versions:
1081
# only doing this when we want multiple versions
1082
# output avoids list copies - which reference and
1083
# dereference many strings.
1084
content = content.copy()
1085
content.apply_delta(delta, version_id)
1086
if multiple_versions:
1087
content_map[component_id] = content
1089
if 'no-eol' in self._index.get_options(version_id):
1090
if multiple_versions:
1091
content = content.copy()
1092
content.strip_last_line_newline()
1093
final_content[version_id] = content
1095
# digest here is the digest from the last applied component.
1096
text = content.text()
1097
actual_sha = sha_strings(text)
1098
if actual_sha != digest:
1099
raise KnitCorrupt(self.filename,
1101
'\n of reconstructed text does not match'
1103
'\n for version %s' %
1104
(actual_sha, digest, version_id))
1105
text_map[version_id] = text
1106
return text_map, final_content
1108
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1110
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1111
if version_ids is None:
1112
version_ids = self.versions()
1114
pb = progress.DummyProgress()
1115
# we don't care about inclusions, the caller cares.
1116
# but we need to setup a list of records to visit.
1117
# we need version_id, position, length
1118
version_id_records = []
1119
requested_versions = set(version_ids)
1120
# filter for available versions
1121
for version_id in requested_versions:
1122
if not self.has_version(version_id):
1123
raise RevisionNotPresent(version_id, self.filename)
1124
# get a in-component-order queue:
1125
for version_id in self.versions():
1126
if version_id in requested_versions:
1127
index_memo = self._index.get_position(version_id)
1128
version_id_records.append((version_id, index_memo))
1130
total = len(version_id_records)
1131
for version_idx, (version_id, data, sha_value) in \
1132
enumerate(self._data.read_records_iter(version_id_records)):
1133
pb.update('Walking content.', version_idx, total)
1134
method = self._index.get_method(version_id)
1136
assert method in ('fulltext', 'line-delta')
1137
if method == 'fulltext':
1138
line_iterator = self.factory.get_fulltext_content(data)
1140
line_iterator = self.factory.get_linedelta_content(data)
1141
# XXX: It might be more efficient to yield (version_id,
1142
# line_iterator) in the future. However for now, this is a simpler
1143
# change to integrate into the rest of the codebase. RBC 20071110
1144
for line in line_iterator:
1145
yield line, version_id
1147
pb.update('Walking content.', total, total)
1149
def iter_parents(self, version_ids):
1150
"""Iterate through the parents for many version ids.
1152
:param version_ids: An iterable yielding version_ids.
1153
:return: An iterator that yields (version_id, parents). Requested
1154
version_ids not present in the versioned file are simply skipped.
1155
The order is undefined, allowing for different optimisations in
1156
the underlying implementation.
1158
return self._index.iter_parents(version_ids)
1160
def num_versions(self):
1161
"""See VersionedFile.num_versions()."""
1162
return self._index.num_versions()
1164
__len__ = num_versions
1166
def annotate_iter(self, version_id):
1167
"""See VersionedFile.annotate_iter."""
1168
return self.factory.annotate_iter(self, version_id)
1170
def get_parents(self, version_id):
1171
"""See VersionedFile.get_parents."""
1174
# 52554 calls in 1264 872 internal down from 3674
1176
return self._index.get_parents(version_id)
1178
raise RevisionNotPresent(version_id, self.filename)
1180
def get_parents_with_ghosts(self, version_id):
1181
"""See VersionedFile.get_parents."""
1183
return self._index.get_parents_with_ghosts(version_id)
1185
raise RevisionNotPresent(version_id, self.filename)
1187
def get_ancestry(self, versions, topo_sorted=True):
1188
"""See VersionedFile.get_ancestry."""
1189
if isinstance(versions, basestring):
1190
versions = [versions]
1193
return self._index.get_ancestry(versions, topo_sorted)
1195
def get_ancestry_with_ghosts(self, versions):
1196
"""See VersionedFile.get_ancestry_with_ghosts."""
1197
if isinstance(versions, basestring):
1198
versions = [versions]
1201
return self._index.get_ancestry_with_ghosts(versions)
1203
def plan_merge(self, ver_a, ver_b):
1204
"""See VersionedFile.plan_merge."""
1205
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1206
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1207
annotated_a = self.annotate(ver_a)
1208
annotated_b = self.annotate(ver_b)
1209
return merge._plan_annotate_merge(annotated_a, annotated_b,
1210
ancestors_a, ancestors_b)
1213
class _KnitComponentFile(object):
1214
"""One of the files used to implement a knit database"""
1216
def __init__(self, transport, filename, mode, file_mode=None,
1217
create_parent_dir=False, dir_mode=None):
1218
self._transport = transport
1219
self._filename = filename
1221
self._file_mode = file_mode
1222
self._dir_mode = dir_mode
1223
self._create_parent_dir = create_parent_dir
1224
self._need_to_create = False
1226
def _full_path(self):
1227
"""Return the full path to this file."""
1228
return self._transport.base + self._filename
1230
def check_header(self, fp):
1231
line = fp.readline()
1233
# An empty file can actually be treated as though the file doesn't
1235
raise errors.NoSuchFile(self._full_path())
1236
if line != self.HEADER:
1237
raise KnitHeaderError(badline=line,
1238
filename=self._transport.abspath(self._filename))
1241
return '%s(%s)' % (self.__class__.__name__, self._filename)
1244
class _KnitIndex(_KnitComponentFile):
1245
"""Manages knit index file.
1247
The index is already kept in memory and read on startup, to enable
1248
fast lookups of revision information. The cursor of the index
1249
file is always pointing to the end, making it easy to append
1252
_cache is a cache for fast mapping from version id to a Index
1255
_history is a cache for fast mapping from indexes to version ids.
1257
The index data format is dictionary compressed when it comes to
1258
parent references; a index entry may only have parents that with a
1259
lover index number. As a result, the index is topological sorted.
1261
Duplicate entries may be written to the index for a single version id
1262
if this is done then the latter one completely replaces the former:
1263
this allows updates to correct version and parent information.
1264
Note that the two entries may share the delta, and that successive
1265
annotations and references MUST point to the first entry.
1267
The index file on disc contains a header, followed by one line per knit
1268
record. The same revision can be present in an index file more than once.
1269
The first occurrence gets assigned a sequence number starting from 0.
1271
The format of a single line is
1272
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1273
REVISION_ID is a utf8-encoded revision id
1274
FLAGS is a comma separated list of flags about the record. Values include
1275
no-eol, line-delta, fulltext.
1276
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1277
that the the compressed data starts at.
1278
LENGTH is the ascii representation of the length of the data file.
1279
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1281
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1282
revision id already in the knit that is a parent of REVISION_ID.
1283
The ' :' marker is the end of record marker.
1286
when a write is interrupted to the index file, it will result in a line
1287
that does not end in ' :'. If the ' :' is not present at the end of a line,
1288
or at the end of the file, then the record that is missing it will be
1289
ignored by the parser.
1291
When writing new records to the index file, the data is preceded by '\n'
1292
to ensure that records always start on new lines even if the last write was
1293
interrupted. As a result its normal for the last line in the index to be
1294
missing a trailing newline. One can be added with no harmful effects.
1297
HEADER = "# bzr knit index 8\n"
1299
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1300
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1302
def _cache_version(self, version_id, options, pos, size, parents):
1303
"""Cache a version record in the history array and index cache.
1305
This is inlined into _load_data for performance. KEEP IN SYNC.
1306
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1309
# only want the _history index to reference the 1st index entry
1311
if version_id not in self._cache:
1312
index = len(self._history)
1313
self._history.append(version_id)
1315
index = self._cache[version_id][5]
1316
self._cache[version_id] = (version_id,
1323
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1324
create_parent_dir=False, delay_create=False, dir_mode=None):
1325
_KnitComponentFile.__init__(self, transport, filename, mode,
1326
file_mode=file_mode,
1327
create_parent_dir=create_parent_dir,
1330
# position in _history is the 'official' index for a revision
1331
# but the values may have come from a newer entry.
1332
# so - wc -l of a knit index is != the number of unique names
1336
fp = self._transport.get(self._filename)
1338
# _load_data may raise NoSuchFile if the target knit is
1340
_load_data(self, fp)
1344
if mode != 'w' or not create:
1347
self._need_to_create = True
1349
self._transport.put_bytes_non_atomic(
1350
self._filename, self.HEADER, mode=self._file_mode)
1352
def get_graph(self):
1353
"""Return a list of the node:parents lists from this knit index."""
1354
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1356
def get_ancestry(self, versions, topo_sorted=True):
1357
"""See VersionedFile.get_ancestry."""
1358
# get a graph of all the mentioned versions:
1360
pending = set(versions)
1363
version = pending.pop()
1366
parents = [p for p in cache[version][4] if p in cache]
1368
raise RevisionNotPresent(version, self._filename)
1369
# if not completed and not a ghost
1370
pending.update([p for p in parents if p not in graph])
1371
graph[version] = parents
1374
return topo_sort(graph.items())
1376
def get_ancestry_with_ghosts(self, versions):
1377
"""See VersionedFile.get_ancestry_with_ghosts."""
1378
# get a graph of all the mentioned versions:
1379
self.check_versions_present(versions)
1382
pending = set(versions)
1384
version = pending.pop()
1386
parents = cache[version][4]
1392
pending.update([p for p in parents if p not in graph])
1393
graph[version] = parents
1394
return topo_sort(graph.items())
1396
def iter_parents(self, version_ids):
1397
"""Iterate through the parents for many version ids.
1399
:param version_ids: An iterable yielding version_ids.
1400
:return: An iterator that yields (version_id, parents). Requested
1401
version_ids not present in the versioned file are simply skipped.
1402
The order is undefined, allowing for different optimisations in
1403
the underlying implementation.
1405
for version_id in version_ids:
1407
yield version_id, tuple(self.get_parents(version_id))
1411
def num_versions(self):
1412
return len(self._history)
1414
__len__ = num_versions
1416
def get_versions(self):
1417
"""Get all the versions in the file. not topologically sorted."""
1418
return self._history
1420
def _version_list_to_index(self, versions):
1423
for version in versions:
1424
if version in cache:
1425
# -- inlined lookup() --
1426
result_list.append(str(cache[version][5]))
1427
# -- end lookup () --
1429
result_list.append('.' + version)
1430
return ' '.join(result_list)
1432
def add_version(self, version_id, options, index_memo, parents):
1433
"""Add a version record to the index."""
1434
self.add_versions(((version_id, options, index_memo, parents),))
1436
def add_versions(self, versions, random_id=False):
1437
"""Add multiple versions to the index.
1439
:param versions: a list of tuples:
1440
(version_id, options, pos, size, parents).
1441
:param random_id: If True the ids being added were randomly generated
1442
and no check for existence will be performed.
1445
orig_history = self._history[:]
1446
orig_cache = self._cache.copy()
1449
for version_id, options, (index, pos, size), parents in versions:
1450
line = "\n%s %s %s %s %s :" % (version_id,
1454
self._version_list_to_index(parents))
1455
assert isinstance(line, str), \
1456
'content must be utf-8 encoded: %r' % (line,)
1458
self._cache_version(version_id, options, pos, size, parents)
1459
if not self._need_to_create:
1460
self._transport.append_bytes(self._filename, ''.join(lines))
1463
sio.write(self.HEADER)
1464
sio.writelines(lines)
1466
self._transport.put_file_non_atomic(self._filename, sio,
1467
create_parent_dir=self._create_parent_dir,
1468
mode=self._file_mode,
1469
dir_mode=self._dir_mode)
1470
self._need_to_create = False
1472
# If any problems happen, restore the original values and re-raise
1473
self._history = orig_history
1474
self._cache = orig_cache
1477
def has_version(self, version_id):
1478
"""True if the version is in the index."""
1479
return version_id in self._cache
1481
def get_position(self, version_id):
1482
"""Return details needed to access the version.
1484
.kndx indices do not support split-out data, so return None for the
1487
:return: a tuple (None, data position, size) to hand to the access
1488
logic to get the record.
1490
entry = self._cache[version_id]
1491
return None, entry[2], entry[3]
1493
def get_method(self, version_id):
1494
"""Return compression method of specified version."""
1496
options = self._cache[version_id][1]
1498
raise RevisionNotPresent(version_id, self._filename)
1499
if 'fulltext' in options:
1502
if 'line-delta' not in options:
1503
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1506
def get_options(self, version_id):
1507
"""Return a string represention options.
1511
return self._cache[version_id][1]
1513
def get_parents(self, version_id):
1514
"""Return parents of specified version ignoring ghosts."""
1515
return [parent for parent in self._cache[version_id][4]
1516
if parent in self._cache]
1518
def get_parents_with_ghosts(self, version_id):
1519
"""Return parents of specified version with ghosts."""
1520
return self._cache[version_id][4]
1522
def check_versions_present(self, version_ids):
1523
"""Check that all specified versions are present."""
1525
for version_id in version_ids:
1526
if version_id not in cache:
1527
raise RevisionNotPresent(version_id, self._filename)
1530
class KnitGraphIndex(object):
1531
"""A knit index that builds on GraphIndex."""
1533
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1534
"""Construct a KnitGraphIndex on a graph_index.
1536
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1537
:param deltas: Allow delta-compressed records.
1538
:param add_callback: If not None, allow additions to the index and call
1539
this callback with a list of added GraphIndex nodes:
1540
[(node, value, node_refs), ...]
1541
:param parents: If True, record knits parents, if not do not record
1544
self._graph_index = graph_index
1545
self._deltas = deltas
1546
self._add_callback = add_callback
1547
self._parents = parents
1548
if deltas and not parents:
1549
raise KnitCorrupt(self, "Cannot do delta compression without "
1552
def _get_entries(self, keys, check_present=False):
1553
"""Get the entries for keys.
1555
:param keys: An iterable of index keys, - 1-tuples.
1560
for node in self._graph_index.iter_entries(keys):
1562
found_keys.add(node[1])
1564
# adapt parentless index to the rest of the code.
1565
for node in self._graph_index.iter_entries(keys):
1566
yield node[0], node[1], node[2], ()
1567
found_keys.add(node[1])
1569
missing_keys = keys.difference(found_keys)
1571
raise RevisionNotPresent(missing_keys.pop(), self)
1573
def _present_keys(self, version_ids):
1575
node[1] for node in self._get_entries(version_ids)])
1577
def _parentless_ancestry(self, versions):
1578
"""Honour the get_ancestry API for parentless knit indices."""
1579
wanted_keys = self._version_ids_to_keys(versions)
1580
present_keys = self._present_keys(wanted_keys)
1581
missing = set(wanted_keys).difference(present_keys)
1583
raise RevisionNotPresent(missing.pop(), self)
1584
return list(self._keys_to_version_ids(present_keys))
1586
def get_ancestry(self, versions, topo_sorted=True):
1587
"""See VersionedFile.get_ancestry."""
1588
if not self._parents:
1589
return self._parentless_ancestry(versions)
1590
# XXX: This will do len(history) index calls - perhaps
1591
# it should be altered to be a index core feature?
1592
# get a graph of all the mentioned versions:
1595
versions = self._version_ids_to_keys(versions)
1596
pending = set(versions)
1598
# get all pending nodes
1599
this_iteration = pending
1600
new_nodes = self._get_entries(this_iteration)
1603
for (index, key, value, node_refs) in new_nodes:
1604
# dont ask for ghosties - otherwise
1605
# we we can end up looping with pending
1606
# being entirely ghosted.
1607
graph[key] = [parent for parent in node_refs[0]
1608
if parent not in ghosts]
1610
for parent in graph[key]:
1611
# dont examine known nodes again
1616
ghosts.update(this_iteration.difference(found))
1617
if versions.difference(graph):
1618
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1620
result_keys = topo_sort(graph.items())
1622
result_keys = graph.iterkeys()
1623
return [key[0] for key in result_keys]
1625
def get_ancestry_with_ghosts(self, versions):
1626
"""See VersionedFile.get_ancestry."""
1627
if not self._parents:
1628
return self._parentless_ancestry(versions)
1629
# XXX: This will do len(history) index calls - perhaps
1630
# it should be altered to be a index core feature?
1631
# get a graph of all the mentioned versions:
1633
versions = self._version_ids_to_keys(versions)
1634
pending = set(versions)
1636
# get all pending nodes
1637
this_iteration = pending
1638
new_nodes = self._get_entries(this_iteration)
1640
for (index, key, value, node_refs) in new_nodes:
1641
graph[key] = node_refs[0]
1643
for parent in graph[key]:
1644
# dont examine known nodes again
1648
missing_versions = this_iteration.difference(graph)
1649
missing_needed = versions.intersection(missing_versions)
1651
raise RevisionNotPresent(missing_needed.pop(), self)
1652
for missing_version in missing_versions:
1653
# add a key, no parents
1654
graph[missing_version] = []
1655
pending.discard(missing_version) # don't look for it
1656
result_keys = topo_sort(graph.items())
1657
return [key[0] for key in result_keys]
1659
def get_graph(self):
1660
"""Return a list of the node:parents lists from this knit index."""
1661
if not self._parents:
1662
return [(key, ()) for key in self.get_versions()]
1664
for index, key, value, refs in self._graph_index.iter_all_entries():
1665
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1668
def iter_parents(self, version_ids):
1669
"""Iterate through the parents for many version ids.
1671
:param version_ids: An iterable yielding version_ids.
1672
:return: An iterator that yields (version_id, parents). Requested
1673
version_ids not present in the versioned file are simply skipped.
1674
The order is undefined, allowing for different optimisations in
1675
the underlying implementation.
1678
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1680
present_parents = set()
1681
for node in all_nodes:
1682
all_parents.update(node[3][0])
1683
# any node we are querying must be present
1684
present_parents.add(node[1])
1685
unknown_parents = all_parents.difference(present_parents)
1686
present_parents.update(self._present_keys(unknown_parents))
1687
for node in all_nodes:
1689
for parent in node[3][0]:
1690
if parent in present_parents:
1691
parents.append(parent[0])
1692
yield node[1][0], tuple(parents)
1694
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1695
yield node[1][0], ()
1697
def num_versions(self):
1698
return len(list(self._graph_index.iter_all_entries()))
1700
__len__ = num_versions
1702
def get_versions(self):
1703
"""Get all the versions in the file. not topologically sorted."""
1704
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1706
def has_version(self, version_id):
1707
"""True if the version is in the index."""
1708
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1710
def _keys_to_version_ids(self, keys):
1711
return tuple(key[0] for key in keys)
1713
def get_position(self, version_id):
1714
"""Return details needed to access the version.
1716
:return: a tuple (index, data position, size) to hand to the access
1717
logic to get the record.
1719
node = self._get_node(version_id)
1720
bits = node[2][1:].split(' ')
1721
return node[0], int(bits[0]), int(bits[1])
1723
def get_method(self, version_id):
1724
"""Return compression method of specified version."""
1725
if not self._deltas:
1727
return self._parent_compression(self._get_node(version_id)[3][1])
1729
def _parent_compression(self, reference_list):
1730
# use the second reference list to decide if this is delta'd or not.
1731
if len(reference_list):
1736
def _get_node(self, version_id):
1738
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1740
raise RevisionNotPresent(version_id, self)
1742
def get_options(self, version_id):
1743
"""Return a string represention options.
1747
node = self._get_node(version_id)
1748
if not self._deltas:
1749
options = ['fulltext']
1751
options = [self._parent_compression(node[3][1])]
1752
if node[2][0] == 'N':
1753
options.append('no-eol')
1756
def get_parents(self, version_id):
1757
"""Return parents of specified version ignoring ghosts."""
1758
parents = list(self.iter_parents([version_id]))
1761
raise errors.RevisionNotPresent(version_id, self)
1762
return parents[0][1]
1764
def get_parents_with_ghosts(self, version_id):
1765
"""Return parents of specified version with ghosts."""
1766
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1767
check_present=True))
1768
if not self._parents:
1770
return self._keys_to_version_ids(nodes[0][3][0])
1772
def check_versions_present(self, version_ids):
1773
"""Check that all specified versions are present."""
1774
keys = self._version_ids_to_keys(version_ids)
1775
present = self._present_keys(keys)
1776
missing = keys.difference(present)
1778
raise RevisionNotPresent(missing.pop(), self)
1780
def add_version(self, version_id, options, access_memo, parents):
1781
"""Add a version record to the index."""
1782
return self.add_versions(((version_id, options, access_memo, parents),))
1784
def add_versions(self, versions, random_id=False):
1785
"""Add multiple versions to the index.
1787
This function does not insert data into the Immutable GraphIndex
1788
backing the KnitGraphIndex, instead it prepares data for insertion by
1789
the caller and checks that it is safe to insert then calls
1790
self._add_callback with the prepared GraphIndex nodes.
1792
:param versions: a list of tuples:
1793
(version_id, options, pos, size, parents).
1794
:param random_id: If True the ids being added were randomly generated
1795
and no check for existence will be performed.
1797
if not self._add_callback:
1798
raise errors.ReadOnlyError(self)
1799
# we hope there are no repositories with inconsistent parentage
1804
for (version_id, options, access_memo, parents) in versions:
1805
index, pos, size = access_memo
1806
key = (version_id, )
1807
parents = tuple((parent, ) for parent in parents)
1808
if 'no-eol' in options:
1812
value += "%d %d" % (pos, size)
1813
if not self._deltas:
1814
if 'line-delta' in options:
1815
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1818
if 'line-delta' in options:
1819
node_refs = (parents, (parents[0],))
1821
node_refs = (parents, ())
1823
node_refs = (parents, )
1826
raise KnitCorrupt(self, "attempt to add node with parents "
1827
"in parentless index.")
1829
keys[key] = (value, node_refs)
1831
present_nodes = self._get_entries(keys)
1832
for (index, key, value, node_refs) in present_nodes:
1833
if (value, node_refs) != keys[key]:
1834
raise KnitCorrupt(self, "inconsistent details in add_versions"
1835
": %s %s" % ((value, node_refs), keys[key]))
1839
for key, (value, node_refs) in keys.iteritems():
1840
result.append((key, value, node_refs))
1842
for key, (value, node_refs) in keys.iteritems():
1843
result.append((key, value))
1844
self._add_callback(result)
1846
def _version_ids_to_keys(self, version_ids):
1847
return set((version_id, ) for version_id in version_ids)
1850
class _KnitAccess(object):
1851
"""Access to knit records in a .knit file."""
1853
def __init__(self, transport, filename, _file_mode, _dir_mode,
1854
_need_to_create, _create_parent_dir):
1855
"""Create a _KnitAccess for accessing and inserting data.
1857
:param transport: The transport the .knit is located on.
1858
:param filename: The filename of the .knit.
1860
self._transport = transport
1861
self._filename = filename
1862
self._file_mode = _file_mode
1863
self._dir_mode = _dir_mode
1864
self._need_to_create = _need_to_create
1865
self._create_parent_dir = _create_parent_dir
1867
def add_raw_records(self, sizes, raw_data):
1868
"""Add raw knit bytes to a storage area.
1870
The data is spooled to whereever the access method is storing data.
1872
:param sizes: An iterable containing the size of each raw data segment.
1873
:param raw_data: A bytestring containing the data.
1874
:return: A list of memos to retrieve the record later. Each memo is a
1875
tuple - (index, pos, length), where the index field is always None
1876
for the .knit access method.
1878
assert type(raw_data) == str, \
1879
'data must be plain bytes was %s' % type(raw_data)
1880
if not self._need_to_create:
1881
base = self._transport.append_bytes(self._filename, raw_data)
1883
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1884
create_parent_dir=self._create_parent_dir,
1885
mode=self._file_mode,
1886
dir_mode=self._dir_mode)
1887
self._need_to_create = False
1891
result.append((None, base, size))
1896
"""IFF this data access has its own storage area, initialise it.
1900
self._transport.put_bytes_non_atomic(self._filename, '',
1901
mode=self._file_mode)
1903
def open_file(self):
1904
"""IFF this data access can be represented as a single file, open it.
1906
For knits that are not mapped to a single file on disk this will
1909
:return: None or a file handle.
1912
return self._transport.get(self._filename)
1917
def get_raw_records(self, memos_for_retrieval):
1918
"""Get the raw bytes for a records.
1920
:param memos_for_retrieval: An iterable containing the (index, pos,
1921
length) memo for retrieving the bytes. The .knit method ignores
1922
the index as there is always only a single file.
1923
:return: An iterator over the bytes of the records.
1925
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1926
for pos, data in self._transport.readv(self._filename, read_vector):
1930
class _PackAccess(object):
1931
"""Access to knit records via a collection of packs."""
1933
def __init__(self, index_to_packs, writer=None):
1934
"""Create a _PackAccess object.
1936
:param index_to_packs: A dict mapping index objects to the transport
1937
and file names for obtaining data.
1938
:param writer: A tuple (pack.ContainerWriter, write_index) which
1939
contains the pack to write, and the index that reads from it will
1943
self.container_writer = writer[0]
1944
self.write_index = writer[1]
1946
self.container_writer = None
1947
self.write_index = None
1948
self.indices = index_to_packs
1950
def add_raw_records(self, sizes, raw_data):
1951
"""Add raw knit bytes to a storage area.
1953
The data is spooled to the container writer in one bytes-record per
1956
:param sizes: An iterable containing the size of each raw data segment.
1957
:param raw_data: A bytestring containing the data.
1958
:return: A list of memos to retrieve the record later. Each memo is a
1959
tuple - (index, pos, length), where the index field is the
1960
write_index object supplied to the PackAccess object.
1962
assert type(raw_data) == str, \
1963
'data must be plain bytes was %s' % type(raw_data)
1967
p_offset, p_length = self.container_writer.add_bytes_record(
1968
raw_data[offset:offset+size], [])
1970
result.append((self.write_index, p_offset, p_length))
1974
"""Pack based knits do not get individually created."""
1976
def get_raw_records(self, memos_for_retrieval):
1977
"""Get the raw bytes for a records.
1979
:param memos_for_retrieval: An iterable containing the (index, pos,
1980
length) memo for retrieving the bytes. The Pack access method
1981
looks up the pack to use for a given record in its index_to_pack
1983
:return: An iterator over the bytes of the records.
1985
# first pass, group into same-index requests
1987
current_index = None
1988
for (index, offset, length) in memos_for_retrieval:
1989
if current_index == index:
1990
current_list.append((offset, length))
1992
if current_index is not None:
1993
request_lists.append((current_index, current_list))
1994
current_index = index
1995
current_list = [(offset, length)]
1996
# handle the last entry
1997
if current_index is not None:
1998
request_lists.append((current_index, current_list))
1999
for index, offsets in request_lists:
2000
transport, path = self.indices[index]
2001
reader = pack.make_readv_reader(transport, path, offsets)
2002
for names, read_func in reader.iter_records():
2003
yield read_func(None)
2005
def open_file(self):
2006
"""Pack based knits have no single file."""
2009
def set_writer(self, writer, index, (transport, packname)):
2010
"""Set a writer to use for adding data."""
2011
if index is not None:
2012
self.indices[index] = (transport, packname)
2013
self.container_writer = writer
2014
self.write_index = index
2017
class _KnitData(object):
2018
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2020
The KnitData class provides the logic for parsing and using knit records,
2021
making use of an access method for the low level read and write operations.
2024
def __init__(self, access):
2025
"""Create a KnitData object.
2027
:param access: The access method to use. Access methods such as
2028
_KnitAccess manage the insertion of raw records and the subsequent
2029
retrieval of the same.
2031
self._access = access
2032
self._checked = False
2033
# TODO: jam 20060713 conceptually, this could spill to disk
2034
# if the cached size gets larger than a certain amount
2035
# but it complicates the model a bit, so for now just use
2036
# a simple dictionary
2038
self._do_cache = False
2040
def enable_cache(self):
2041
"""Enable caching of reads."""
2042
self._do_cache = True
2044
def clear_cache(self):
2045
"""Clear the record cache."""
2046
self._do_cache = False
2049
def _open_file(self):
2050
return self._access.open_file()
2052
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2053
"""Convert version_id, digest, lines into a raw data block.
2055
:param dense_lines: The bytes of lines but in a denser form. For
2056
instance, if lines is a list of 1000 bytestrings each ending in \n,
2057
dense_lines may be a list with one line in it, containing all the
2058
1000's lines and their \n's. Using dense_lines if it is already
2059
known is a win because the string join to create bytes in this
2060
function spends less time resizing the final string.
2061
:return: (len, a StringIO instance with the raw data ready to read.)
2063
# Note: using a string copy here increases memory pressure with e.g.
2064
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2065
# when doing the initial commit of a mozilla tree. RBC 20070921
2066
bytes = ''.join(chain(
2067
["version %s %d %s\n" % (version_id,
2070
dense_lines or lines,
2071
["end %s\n" % version_id]))
2072
assert bytes.__class__ == str
2073
compressed_bytes = bytes_to_gzip(bytes)
2074
return len(compressed_bytes), compressed_bytes
2076
def add_raw_records(self, sizes, raw_data):
2077
"""Append a prepared record to the data file.
2079
:param sizes: An iterable containing the size of each raw data segment.
2080
:param raw_data: A bytestring containing the data.
2081
:return: a list of index data for the way the data was stored.
2082
See the access method add_raw_records documentation for more
2085
return self._access.add_raw_records(sizes, raw_data)
2087
def _parse_record_header(self, version_id, raw_data):
2088
"""Parse a record header for consistency.
2090
:return: the header and the decompressor stream.
2091
as (stream, header_record)
2093
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2095
rec = self._check_header(version_id, df.readline())
2096
except Exception, e:
2097
raise KnitCorrupt(self._access,
2098
"While reading {%s} got %s(%s)"
2099
% (version_id, e.__class__.__name__, str(e)))
2102
def _check_header(self, version_id, line):
2105
raise KnitCorrupt(self._access,
2106
'unexpected number of elements in record header')
2107
if rec[1] != version_id:
2108
raise KnitCorrupt(self._access,
2109
'unexpected version, wanted %r, got %r'
2110
% (version_id, rec[1]))
2113
def _parse_record(self, version_id, data):
2115
# 4168 calls in 2880 217 internal
2116
# 4168 calls to _parse_record_header in 2121
2117
# 4168 calls to readlines in 330
2118
df = GzipFile(mode='rb', fileobj=StringIO(data))
2121
record_contents = df.readlines()
2122
except Exception, e:
2123
raise KnitCorrupt(self._access,
2124
"While reading {%s} got %s(%s)"
2125
% (version_id, e.__class__.__name__, str(e)))
2126
header = record_contents.pop(0)
2127
rec = self._check_header(version_id, header)
2129
last_line = record_contents.pop()
2130
if len(record_contents) != int(rec[2]):
2131
raise KnitCorrupt(self._access,
2132
'incorrect number of lines %s != %s'
2134
% (len(record_contents), int(rec[2]),
2136
if last_line != 'end %s\n' % rec[1]:
2137
raise KnitCorrupt(self._access,
2138
'unexpected version end line %r, wanted %r'
2139
% (last_line, version_id))
2141
return record_contents, rec[3]
2143
def read_records_iter_raw(self, records):
2144
"""Read text records from data file and yield raw data.
2146
This unpacks enough of the text record to validate the id is
2147
as expected but thats all.
2149
# setup an iterator of the external records:
2150
# uses readv so nice and fast we hope.
2152
# grab the disk data needed.
2154
# Don't check _cache if it is empty
2155
needed_offsets = [index_memo for version_id, index_memo
2157
if version_id not in self._cache]
2159
needed_offsets = [index_memo for version_id, index_memo
2162
raw_records = self._access.get_raw_records(needed_offsets)
2164
for version_id, index_memo in records:
2165
if version_id in self._cache:
2166
# This data has already been validated
2167
data = self._cache[version_id]
2169
data = raw_records.next()
2171
self._cache[version_id] = data
2173
# validate the header
2174
df, rec = self._parse_record_header(version_id, data)
2176
yield version_id, data
2178
def read_records_iter(self, records):
2179
"""Read text records from data file and yield result.
2181
The result will be returned in whatever is the fastest to read.
2182
Not by the order requested. Also, multiple requests for the same
2183
record will only yield 1 response.
2184
:param records: A list of (version_id, pos, len) entries
2185
:return: Yields (version_id, contents, digest) in the order
2186
read, not the order requested
2192
# Skip records we have alread seen
2193
yielded_records = set()
2194
needed_records = set()
2195
for record in records:
2196
if record[0] in self._cache:
2197
if record[0] in yielded_records:
2199
yielded_records.add(record[0])
2200
data = self._cache[record[0]]
2201
content, digest = self._parse_record(record[0], data)
2202
yield (record[0], content, digest)
2204
needed_records.add(record)
2205
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2207
needed_records = sorted(set(records), key=operator.itemgetter(1))
2209
if not needed_records:
2212
# The transport optimizes the fetching as well
2213
# (ie, reads continuous ranges.)
2214
raw_data = self._access.get_raw_records(
2215
[index_memo for version_id, index_memo in needed_records])
2217
for (version_id, index_memo), data in \
2218
izip(iter(needed_records), raw_data):
2219
content, digest = self._parse_record(version_id, data)
2221
self._cache[version_id] = data
2222
yield version_id, content, digest
2224
def read_records(self, records):
2225
"""Read records into a dictionary."""
2227
for record_id, content, digest in \
2228
self.read_records_iter(records):
2229
components[record_id] = (content, digest)
2233
class InterKnit(InterVersionedFile):
2234
"""Optimised code paths for knit to knit operations."""
2236
_matching_file_from_factory = KnitVersionedFile
2237
_matching_file_to_factory = KnitVersionedFile
2240
def is_compatible(source, target):
2241
"""Be compatible with knits. """
2243
return (isinstance(source, KnitVersionedFile) and
2244
isinstance(target, KnitVersionedFile))
2245
except AttributeError:
2248
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2249
"""Copy texts to the target by extracting and adding them one by one.
2251
see join() for the parameter definitions.
2253
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2254
graph = self.source.get_graph(version_ids)
2255
order = topo_sort(graph.items())
2257
def size_of_content(content):
2258
return sum(len(line) for line in content.text())
2259
# Cache at most 10MB of parent texts
2260
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2261
compute_size=size_of_content)
2262
# TODO: jam 20071116 It would be nice to have a streaming interface to
2263
# get multiple texts from a source. The source could be smarter
2264
# about how it handled intermediate stages.
2265
# get_line_list() or make_mpdiffs() seem like a possibility, but
2266
# at the moment they extract all full texts into memory, which
2267
# causes us to store more than our 3x fulltext goal.
2268
# Repository.iter_files_bytes() may be another possibility
2269
to_process = [version for version in order
2270
if version not in self.target]
2271
total = len(to_process)
2272
pb = ui.ui_factory.nested_progress_bar()
2274
for index, version in enumerate(to_process):
2275
pb.update('Converting versioned data', index, total)
2276
sha1, num_bytes, parent_text = self.target.add_lines(version,
2277
self.source.get_parents(version),
2278
self.source.get_lines(version),
2279
parent_texts=parent_cache)
2280
parent_cache[version] = parent_text
2285
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2286
"""See InterVersionedFile.join."""
2287
assert isinstance(self.source, KnitVersionedFile)
2288
assert isinstance(self.target, KnitVersionedFile)
2290
# If the source and target are mismatched w.r.t. annotations vs
2291
# plain, the data needs to be converted accordingly
2292
if self.source.factory.annotated == self.target.factory.annotated:
2294
elif self.source.factory.annotated:
2295
converter = self._anno_to_plain_converter
2297
# We're converting from a plain to an annotated knit. Copy them
2298
# across by full texts.
2299
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2301
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2305
pb = ui.ui_factory.nested_progress_bar()
2307
version_ids = list(version_ids)
2308
if None in version_ids:
2309
version_ids.remove(None)
2311
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2312
this_versions = set(self.target._index.get_versions())
2313
# XXX: For efficiency we should not look at the whole index,
2314
# we only need to consider the referenced revisions - they
2315
# must all be present, or the method must be full-text.
2316
# TODO, RBC 20070919
2317
needed_versions = self.source_ancestry - this_versions
2319
if not needed_versions:
2321
full_list = topo_sort(self.source.get_graph())
2323
version_list = [i for i in full_list if (not self.target.has_version(i)
2324
and i in needed_versions)]
2328
copy_queue_records = []
2330
for version_id in version_list:
2331
options = self.source._index.get_options(version_id)
2332
parents = self.source._index.get_parents_with_ghosts(version_id)
2333
# check that its will be a consistent copy:
2334
for parent in parents:
2335
# if source has the parent, we must :
2336
# * already have it or
2337
# * have it scheduled already
2338
# otherwise we don't care
2339
assert (self.target.has_version(parent) or
2340
parent in copy_set or
2341
not self.source.has_version(parent))
2342
index_memo = self.source._index.get_position(version_id)
2343
copy_queue_records.append((version_id, index_memo))
2344
copy_queue.append((version_id, options, parents))
2345
copy_set.add(version_id)
2347
# data suck the join:
2349
total = len(version_list)
2352
for (version_id, raw_data), \
2353
(version_id2, options, parents) in \
2354
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2356
assert version_id == version_id2, 'logic error, inconsistent results'
2358
pb.update("Joining knit", count, total)
2360
size, raw_data = converter(raw_data, version_id, options,
2363
size = len(raw_data)
2364
raw_records.append((version_id, options, parents, size))
2365
raw_datum.append(raw_data)
2366
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2371
def _anno_to_plain_converter(self, raw_data, version_id, options,
2373
"""Convert annotated content to plain content."""
2374
data, digest = self.source._data._parse_record(version_id, raw_data)
2375
if 'fulltext' in options:
2376
content = self.source.factory.parse_fulltext(data, version_id)
2377
lines = self.target.factory.lower_fulltext(content)
2379
delta = self.source.factory.parse_line_delta(data, version_id,
2381
lines = self.target.factory.lower_line_delta(delta)
2382
return self.target._data._record_to_data(version_id, digest, lines)
2385
InterVersionedFile.register_optimiser(InterKnit)
2388
class WeaveToKnit(InterVersionedFile):
2389
"""Optimised code paths for weave to knit operations."""
2391
_matching_file_from_factory = bzrlib.weave.WeaveFile
2392
_matching_file_to_factory = KnitVersionedFile
2395
def is_compatible(source, target):
2396
"""Be compatible with weaves to knits."""
2398
return (isinstance(source, bzrlib.weave.Weave) and
2399
isinstance(target, KnitVersionedFile))
2400
except AttributeError:
2403
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2404
"""See InterVersionedFile.join."""
2405
assert isinstance(self.source, bzrlib.weave.Weave)
2406
assert isinstance(self.target, KnitVersionedFile)
2408
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2413
pb = ui.ui_factory.nested_progress_bar()
2415
version_ids = list(version_ids)
2417
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2418
this_versions = set(self.target._index.get_versions())
2419
needed_versions = self.source_ancestry - this_versions
2421
if not needed_versions:
2423
full_list = topo_sort(self.source.get_graph())
2425
version_list = [i for i in full_list if (not self.target.has_version(i)
2426
and i in needed_versions)]
2430
total = len(version_list)
2431
for version_id in version_list:
2432
pb.update("Converting to knit", count, total)
2433
parents = self.source.get_parents(version_id)
2434
# check that its will be a consistent copy:
2435
for parent in parents:
2436
# if source has the parent, we must already have it
2437
assert (self.target.has_version(parent))
2438
self.target.add_lines(
2439
version_id, parents, self.source.get_lines(version_id))
2446
InterVersionedFile.register_optimiser(WeaveToKnit)
2449
# Deprecated, use PatienceSequenceMatcher instead
2450
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2453
def annotate_knit(knit, revision_id):
2454
"""Annotate a knit with no cached annotations.
2456
This implementation is for knits with no cached annotations.
2457
It will work for knits with cached annotations, but this is not
2460
ancestry = knit.get_ancestry(revision_id)
2461
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2463
for candidate in ancestry:
2464
if candidate in annotations:
2466
parents = knit.get_parents(candidate)
2467
if len(parents) == 0:
2469
elif knit._index.get_method(candidate) != 'line-delta':
2472
parent, sha1, noeol, delta = knit.get_delta(candidate)
2473
blocks = KnitContent.get_line_delta_blocks(delta,
2474
fulltext[parents[0]], fulltext[candidate])
2475
annotations[candidate] = list(annotate.reannotate([annotations[p]
2476
for p in parents], fulltext[candidate], candidate, blocks))
2477
return iter(annotations[revision_id])
2481
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2483
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
