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
if not self.has_version(version_id):
603
raise RevisionNotPresent(version_id, self.filename)
604
options = self._index.get_options(version_id)
605
parents = self._index.get_parents_with_ghosts(version_id)
606
index_memo = self._index.get_position(version_id)
607
version_index[version_id] = (index_memo, options, parents)
608
if parents and parents[0] in required_version_set:
609
# must wait until the parent has been sent
610
deferred.setdefault(parents[0], []). \
613
# either a fulltext, or a delta whose parent the client did
614
# not ask for and presumably already has
615
ready_to_send.append(version_id)
616
# build a list of results to return, plus instructions for data to
618
copy_queue_records = []
619
temp_version_list = []
621
# XXX: pushing and popping lists may be a bit inefficient
622
version_id = ready_to_send.pop(0)
623
(index_memo, options, parents) = version_index[version_id]
624
copy_queue_records.append((version_id, index_memo))
625
none, data_pos, data_size = index_memo
626
temp_version_list.append((version_id, options, data_size,
628
if version_id in deferred:
629
# now we can send all the children of this revision - we could
630
# put them in anywhere, but we hope that sending them soon
631
# after the fulltext will give good locality in the receiver
632
ready_to_send[:0] = deferred.pop(version_id)
633
assert len(deferred) == 0, \
634
"Still have compressed child versions waiting to be sent"
635
# XXX: The stream format is such that we cannot stream it - we have to
636
# know the length of all the data a-priori.
638
result_version_list = []
639
for (version_id, raw_data), \
640
(version_id2, options, _, parents) in \
641
izip(self._data.read_records_iter_raw(copy_queue_records),
643
assert version_id == version_id2, \
644
'logic error, inconsistent results'
645
raw_datum.append(raw_data)
646
result_version_list.append(
647
(version_id, options, len(raw_data), parents))
648
# provide a callback to get data incrementally.
649
pseudo_file = StringIO(''.join(raw_datum))
652
return pseudo_file.read()
654
return pseudo_file.read(length)
655
return (self.get_format_signature(), result_version_list, read)
657
def _extract_blocks(self, version_id, source, target):
658
if self._index.get_method(version_id) != 'line-delta':
660
parent, sha1, noeol, delta = self.get_delta(version_id)
661
return KnitContent.get_line_delta_blocks(delta, source, target)
663
def get_delta(self, version_id):
664
"""Get a delta for constructing version from some other version."""
665
self.check_not_reserved_id(version_id)
666
parents = self.get_parents(version_id)
671
index_memo = self._index.get_position(version_id)
672
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
673
noeol = 'no-eol' in self._index.get_options(version_id)
674
if 'fulltext' == self._index.get_method(version_id):
675
new_content = self.factory.parse_fulltext(data, version_id)
676
if parent is not None:
677
reference_content = self._get_content(parent)
678
old_texts = reference_content.text()
681
new_texts = new_content.text()
682
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
684
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
686
delta = self.factory.parse_line_delta(data, version_id)
687
return parent, sha1, noeol, delta
689
def get_format_signature(self):
690
"""See VersionedFile.get_format_signature()."""
691
if self.factory.annotated:
692
annotated_part = "annotated"
694
annotated_part = "plain"
695
return "knit-%s" % (annotated_part,)
697
def get_graph_with_ghosts(self):
698
"""See VersionedFile.get_graph_with_ghosts()."""
699
graph_items = self._index.get_graph()
700
return dict(graph_items)
702
def get_sha1(self, version_id):
703
return self.get_sha1s([version_id])[0]
705
def get_sha1s(self, version_ids):
706
"""See VersionedFile.get_sha1()."""
707
record_map = self._get_record_map(version_ids)
708
# record entry 2 is the 'digest'.
709
return [record_map[v][2] for v in version_ids]
713
"""See VersionedFile.get_suffixes()."""
714
return [DATA_SUFFIX, INDEX_SUFFIX]
716
def has_ghost(self, version_id):
717
"""True if there is a ghost reference in the file to version_id."""
719
if self.has_version(version_id):
721
# optimisable if needed by memoising the _ghosts set.
722
items = self._index.get_graph()
723
for node, parents in items:
724
for parent in parents:
725
if parent not in self._index._cache:
726
if parent == version_id:
730
def insert_data_stream(self, (format, data_list, reader_callable)):
731
"""Insert knit records from a data stream into this knit.
733
If a version in the stream is already present in this knit, it will not
734
be inserted a second time. It will be checked for consistency with the
735
stored version however, and may cause a KnitCorrupt error to be raised
736
if the data in the stream disagrees with the already stored data.
738
:seealso: get_data_stream
740
if format != self.get_format_signature():
741
trace.mutter('incompatible format signature inserting to %r', self)
742
raise KnitDataStreamIncompatible(
743
format, self.get_format_signature())
745
for version_id, options, length, parents in data_list:
746
if self.has_version(version_id):
747
# First check: the list of parents.
748
my_parents = self.get_parents_with_ghosts(version_id)
749
if my_parents != parents:
750
# XXX: KnitCorrupt is not quite the right exception here.
753
'parents list %r from data stream does not match '
754
'already recorded parents %r for %s'
755
% (parents, my_parents, version_id))
757
# Also check the SHA-1 of the fulltext this content will
759
raw_data = reader_callable(length)
760
my_fulltext_sha1 = self.get_sha1(version_id)
761
df, rec = self._data._parse_record_header(version_id, raw_data)
762
stream_fulltext_sha1 = rec[3]
763
if my_fulltext_sha1 != stream_fulltext_sha1:
764
# Actually, we don't know if it's this knit that's corrupt,
765
# or the data stream we're trying to insert.
767
self.filename, 'sha-1 does not match %s' % version_id)
769
if 'line-delta' in options:
770
# Make sure that this knit record is actually useful: a
771
# line-delta is no use unless we have its parent.
772
# Fetching from a broken repository with this problem
773
# shouldn't break the target repository.
774
if not self._index.has_version(parents[0]):
777
'line-delta from stream references '
778
'missing parent %s' % parents[0])
779
self._add_raw_records(
780
[(version_id, options, parents, length)],
781
reader_callable(length))
784
"""See VersionedFile.versions."""
785
if 'evil' in debug.debug_flags:
786
trace.mutter_callsite(2, "versions scales with size of history")
787
return self._index.get_versions()
789
def has_version(self, version_id):
790
"""See VersionedFile.has_version."""
791
if 'evil' in debug.debug_flags:
792
trace.mutter_callsite(2, "has_version is a LBYL scenario")
793
return self._index.has_version(version_id)
795
__contains__ = has_version
797
def _merge_annotations(self, content, parents, parent_texts={},
798
delta=None, annotated=None,
799
left_matching_blocks=None):
800
"""Merge annotations for content. This is done by comparing
801
the annotations based on changed to the text.
803
if left_matching_blocks is not None:
804
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
808
for parent_id in parents:
809
merge_content = self._get_content(parent_id, parent_texts)
810
if (parent_id == parents[0] and delta_seq is not None):
813
seq = patiencediff.PatienceSequenceMatcher(
814
None, merge_content.text(), content.text())
815
for i, j, n in seq.get_matching_blocks():
818
# this appears to copy (origin, text) pairs across to the
819
# new content for any line that matches the last-checked
821
content._lines[j:j+n] = merge_content._lines[i:i+n]
823
if delta_seq is None:
824
reference_content = self._get_content(parents[0], parent_texts)
825
new_texts = content.text()
826
old_texts = reference_content.text()
827
delta_seq = patiencediff.PatienceSequenceMatcher(
828
None, old_texts, new_texts)
829
return self._make_line_delta(delta_seq, content)
831
def _make_line_delta(self, delta_seq, new_content):
832
"""Generate a line delta from delta_seq and new_content."""
834
for op in delta_seq.get_opcodes():
837
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
840
def _get_components_positions(self, version_ids):
841
"""Produce a map of position data for the components of versions.
843
This data is intended to be used for retrieving the knit records.
845
A dict of version_id to (method, data_pos, data_size, next) is
847
method is the way referenced data should be applied.
848
data_pos is the position of the data in the knit.
849
data_size is the size of the data in the knit.
850
next is the build-parent of the version, or None for fulltexts.
853
for version_id in version_ids:
856
while cursor is not None and cursor not in component_data:
857
method = self._index.get_method(cursor)
858
if method == 'fulltext':
861
next = self.get_parents_with_ghosts(cursor)[0]
862
index_memo = self._index.get_position(cursor)
863
component_data[cursor] = (method, index_memo, next)
865
return component_data
867
def _get_content(self, version_id, parent_texts={}):
868
"""Returns a content object that makes up the specified
870
cached_version = parent_texts.get(version_id, None)
871
if cached_version is not None:
872
if not self.has_version(version_id):
873
raise RevisionNotPresent(version_id, self.filename)
874
return cached_version
876
text_map, contents_map = self._get_content_maps([version_id])
877
return contents_map[version_id]
879
def _check_versions_present(self, version_ids):
880
"""Check that all specified versions are present."""
881
self._index.check_versions_present(version_ids)
883
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
884
nostore_sha, random_id, check_content):
885
"""See VersionedFile.add_lines_with_ghosts()."""
886
self._check_add(version_id, lines, random_id, check_content)
887
return self._add(version_id, lines, parents, self.delta,
888
parent_texts, None, nostore_sha, random_id)
890
def _add_lines(self, version_id, parents, lines, parent_texts,
891
left_matching_blocks, nostore_sha, random_id, check_content):
892
"""See VersionedFile.add_lines."""
893
self._check_add(version_id, lines, random_id, check_content)
894
self._check_versions_present(parents)
895
return self._add(version_id, lines[:], parents, self.delta,
896
parent_texts, left_matching_blocks, nostore_sha, random_id)
898
def _check_add(self, version_id, lines, random_id, check_content):
899
"""check that version_id and lines are safe to add."""
900
if contains_whitespace(version_id):
901
raise InvalidRevisionId(version_id, self.filename)
902
self.check_not_reserved_id(version_id)
903
# Technically this could be avoided if we are happy to allow duplicate
904
# id insertion when other things than bzr core insert texts, but it
905
# seems useful for folk using the knit api directly to have some safety
906
# blanket that we can disable.
907
if not random_id and self.has_version(version_id):
908
raise RevisionAlreadyPresent(version_id, self.filename)
910
self._check_lines_not_unicode(lines)
911
self._check_lines_are_lines(lines)
913
def _add(self, version_id, lines, parents, delta, parent_texts,
914
left_matching_blocks, nostore_sha, random_id):
915
"""Add a set of lines on top of version specified by parents.
917
If delta is true, compress the text as a line-delta against
920
Any versions not present will be converted into ghosts.
922
# first thing, if the content is something we don't need to store, find
924
line_bytes = ''.join(lines)
925
digest = sha_string(line_bytes)
926
if nostore_sha == digest:
927
raise errors.ExistingContent
930
if parent_texts is None:
932
for parent in parents:
933
if self.has_version(parent):
934
present_parents.append(parent)
936
# can only compress against the left most present parent.
938
(len(present_parents) == 0 or
939
present_parents[0] != parents[0])):
942
text_length = len(line_bytes)
945
if lines[-1][-1] != '\n':
946
# copy the contents of lines.
948
options.append('no-eol')
949
lines[-1] = lines[-1] + '\n'
953
# To speed the extract of texts the delta chain is limited
954
# to a fixed number of deltas. This should minimize both
955
# I/O and the time spend applying deltas.
956
delta = self._check_should_delta(present_parents)
958
assert isinstance(version_id, str)
959
content = self.factory.make(lines, version_id)
960
if delta or (self.factory.annotated and len(present_parents) > 0):
961
# Merge annotations from parent texts if needed.
962
delta_hunks = self._merge_annotations(content, present_parents,
963
parent_texts, delta, self.factory.annotated,
964
left_matching_blocks)
967
options.append('line-delta')
968
store_lines = self.factory.lower_line_delta(delta_hunks)
969
size, bytes = self._data._record_to_data(version_id, digest,
972
options.append('fulltext')
973
# isinstance is slower and we have no hierarchy.
974
if self.factory.__class__ == KnitPlainFactory:
975
# Use the already joined bytes saving iteration time in
977
size, bytes = self._data._record_to_data(version_id, digest,
980
# get mixed annotation + content and feed it into the
982
store_lines = self.factory.lower_fulltext(content)
983
size, bytes = self._data._record_to_data(version_id, digest,
986
access_memo = self._data.add_raw_records([size], bytes)[0]
987
self._index.add_versions(
988
((version_id, options, access_memo, parents),),
990
return digest, text_length, content
992
def check(self, progress_bar=None):
993
"""See VersionedFile.check()."""
995
def _clone_text(self, new_version_id, old_version_id, parents):
996
"""See VersionedFile.clone_text()."""
997
# FIXME RBC 20060228 make fast by only inserting an index with null
999
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1001
def get_lines(self, version_id):
1002
"""See VersionedFile.get_lines()."""
1003
return self.get_line_list([version_id])[0]
1005
def _get_record_map(self, version_ids):
1006
"""Produce a dictionary of knit records.
1008
The keys are version_ids, the values are tuples of (method, content,
1010
method is the way the content should be applied.
1011
content is a KnitContent object.
1012
digest is the SHA1 digest of this version id after all steps are done
1013
next is the build-parent of the version, i.e. the leftmost ancestor.
1014
If the method is fulltext, next will be None.
1016
position_map = self._get_components_positions(version_ids)
1017
# c = component_id, m = method, i_m = index_memo, n = next
1018
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1020
for component_id, content, digest in \
1021
self._data.read_records_iter(records):
1022
method, index_memo, next = position_map[component_id]
1023
record_map[component_id] = method, content, digest, next
1027
def get_text(self, version_id):
1028
"""See VersionedFile.get_text"""
1029
return self.get_texts([version_id])[0]
1031
def get_texts(self, version_ids):
1032
return [''.join(l) for l in self.get_line_list(version_ids)]
1034
def get_line_list(self, version_ids):
1035
"""Return the texts of listed versions as a list of strings."""
1036
for version_id in version_ids:
1037
self.check_not_reserved_id(version_id)
1038
text_map, content_map = self._get_content_maps(version_ids)
1039
return [text_map[v] for v in version_ids]
1041
_get_lf_split_line_list = get_line_list
1043
def _get_content_maps(self, version_ids):
1044
"""Produce maps of text and KnitContents
1046
:return: (text_map, content_map) where text_map contains the texts for
1047
the requested versions and content_map contains the KnitContents.
1048
Both dicts take version_ids as their keys.
1050
# FUTURE: This function could be improved for the 'extract many' case
1051
# by tracking each component and only doing the copy when the number of
1052
# children than need to apply delta's to it is > 1 or it is part of the
1054
version_ids = list(version_ids)
1055
multiple_versions = len(version_ids) != 1
1056
record_map = self._get_record_map(version_ids)
1061
for version_id in version_ids:
1064
while cursor is not None:
1065
method, data, digest, next = record_map[cursor]
1066
components.append((cursor, method, data, digest))
1067
if cursor in content_map:
1072
for component_id, method, data, digest in reversed(components):
1073
if component_id in content_map:
1074
content = content_map[component_id]
1076
if method == 'fulltext':
1077
assert content is None
1078
content = self.factory.parse_fulltext(data, version_id)
1079
elif method == 'line-delta':
1080
delta = self.factory.parse_line_delta(data, version_id)
1081
if multiple_versions:
1082
# only doing this when we want multiple versions
1083
# output avoids list copies - which reference and
1084
# dereference many strings.
1085
content = content.copy()
1086
content.apply_delta(delta, version_id)
1087
if multiple_versions:
1088
content_map[component_id] = content
1090
if 'no-eol' in self._index.get_options(version_id):
1091
if multiple_versions:
1092
content = content.copy()
1093
content.strip_last_line_newline()
1094
final_content[version_id] = content
1096
# digest here is the digest from the last applied component.
1097
text = content.text()
1098
actual_sha = sha_strings(text)
1099
if actual_sha != digest:
1100
raise KnitCorrupt(self.filename,
1102
'\n of reconstructed text does not match'
1104
'\n for version %s' %
1105
(actual_sha, digest, version_id))
1106
text_map[version_id] = text
1107
return text_map, final_content
1109
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1111
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1112
if version_ids is None:
1113
version_ids = self.versions()
1115
pb = progress.DummyProgress()
1116
# we don't care about inclusions, the caller cares.
1117
# but we need to setup a list of records to visit.
1118
# we need version_id, position, length
1119
version_id_records = []
1120
requested_versions = set(version_ids)
1121
# filter for available versions
1122
for version_id in requested_versions:
1123
if not self.has_version(version_id):
1124
raise RevisionNotPresent(version_id, self.filename)
1125
# get a in-component-order queue:
1126
for version_id in self.versions():
1127
if version_id in requested_versions:
1128
index_memo = self._index.get_position(version_id)
1129
version_id_records.append((version_id, index_memo))
1131
total = len(version_id_records)
1132
for version_idx, (version_id, data, sha_value) in \
1133
enumerate(self._data.read_records_iter(version_id_records)):
1134
pb.update('Walking content.', version_idx, total)
1135
method = self._index.get_method(version_id)
1137
assert method in ('fulltext', 'line-delta')
1138
if method == 'fulltext':
1139
line_iterator = self.factory.get_fulltext_content(data)
1141
line_iterator = self.factory.get_linedelta_content(data)
1142
# XXX: It might be more efficient to yield (version_id,
1143
# line_iterator) in the future. However for now, this is a simpler
1144
# change to integrate into the rest of the codebase. RBC 20071110
1145
for line in line_iterator:
1146
yield line, version_id
1148
pb.update('Walking content.', total, total)
1150
def iter_parents(self, version_ids):
1151
"""Iterate through the parents for many version ids.
1153
:param version_ids: An iterable yielding version_ids.
1154
:return: An iterator that yields (version_id, parents). Requested
1155
version_ids not present in the versioned file are simply skipped.
1156
The order is undefined, allowing for different optimisations in
1157
the underlying implementation.
1159
return self._index.iter_parents(version_ids)
1161
def num_versions(self):
1162
"""See VersionedFile.num_versions()."""
1163
return self._index.num_versions()
1165
__len__ = num_versions
1167
def annotate_iter(self, version_id):
1168
"""See VersionedFile.annotate_iter."""
1169
return self.factory.annotate_iter(self, version_id)
1171
def get_parents(self, version_id):
1172
"""See VersionedFile.get_parents."""
1175
# 52554 calls in 1264 872 internal down from 3674
1177
return self._index.get_parents(version_id)
1179
raise RevisionNotPresent(version_id, self.filename)
1181
def get_parents_with_ghosts(self, version_id):
1182
"""See VersionedFile.get_parents."""
1184
return self._index.get_parents_with_ghosts(version_id)
1186
raise RevisionNotPresent(version_id, self.filename)
1188
def get_ancestry(self, versions, topo_sorted=True):
1189
"""See VersionedFile.get_ancestry."""
1190
if isinstance(versions, basestring):
1191
versions = [versions]
1194
return self._index.get_ancestry(versions, topo_sorted)
1196
def get_ancestry_with_ghosts(self, versions):
1197
"""See VersionedFile.get_ancestry_with_ghosts."""
1198
if isinstance(versions, basestring):
1199
versions = [versions]
1202
return self._index.get_ancestry_with_ghosts(versions)
1204
def plan_merge(self, ver_a, ver_b):
1205
"""See VersionedFile.plan_merge."""
1206
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1207
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1208
annotated_a = self.annotate(ver_a)
1209
annotated_b = self.annotate(ver_b)
1210
return merge._plan_annotate_merge(annotated_a, annotated_b,
1211
ancestors_a, ancestors_b)
1214
class _KnitComponentFile(object):
1215
"""One of the files used to implement a knit database"""
1217
def __init__(self, transport, filename, mode, file_mode=None,
1218
create_parent_dir=False, dir_mode=None):
1219
self._transport = transport
1220
self._filename = filename
1222
self._file_mode = file_mode
1223
self._dir_mode = dir_mode
1224
self._create_parent_dir = create_parent_dir
1225
self._need_to_create = False
1227
def _full_path(self):
1228
"""Return the full path to this file."""
1229
return self._transport.base + self._filename
1231
def check_header(self, fp):
1232
line = fp.readline()
1234
# An empty file can actually be treated as though the file doesn't
1236
raise errors.NoSuchFile(self._full_path())
1237
if line != self.HEADER:
1238
raise KnitHeaderError(badline=line,
1239
filename=self._transport.abspath(self._filename))
1242
return '%s(%s)' % (self.__class__.__name__, self._filename)
1245
class _KnitIndex(_KnitComponentFile):
1246
"""Manages knit index file.
1248
The index is already kept in memory and read on startup, to enable
1249
fast lookups of revision information. The cursor of the index
1250
file is always pointing to the end, making it easy to append
1253
_cache is a cache for fast mapping from version id to a Index
1256
_history is a cache for fast mapping from indexes to version ids.
1258
The index data format is dictionary compressed when it comes to
1259
parent references; a index entry may only have parents that with a
1260
lover index number. As a result, the index is topological sorted.
1262
Duplicate entries may be written to the index for a single version id
1263
if this is done then the latter one completely replaces the former:
1264
this allows updates to correct version and parent information.
1265
Note that the two entries may share the delta, and that successive
1266
annotations and references MUST point to the first entry.
1268
The index file on disc contains a header, followed by one line per knit
1269
record. The same revision can be present in an index file more than once.
1270
The first occurrence gets assigned a sequence number starting from 0.
1272
The format of a single line is
1273
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1274
REVISION_ID is a utf8-encoded revision id
1275
FLAGS is a comma separated list of flags about the record. Values include
1276
no-eol, line-delta, fulltext.
1277
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1278
that the the compressed data starts at.
1279
LENGTH is the ascii representation of the length of the data file.
1280
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1282
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1283
revision id already in the knit that is a parent of REVISION_ID.
1284
The ' :' marker is the end of record marker.
1287
when a write is interrupted to the index file, it will result in a line
1288
that does not end in ' :'. If the ' :' is not present at the end of a line,
1289
or at the end of the file, then the record that is missing it will be
1290
ignored by the parser.
1292
When writing new records to the index file, the data is preceded by '\n'
1293
to ensure that records always start on new lines even if the last write was
1294
interrupted. As a result its normal for the last line in the index to be
1295
missing a trailing newline. One can be added with no harmful effects.
1298
HEADER = "# bzr knit index 8\n"
1300
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1301
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1303
def _cache_version(self, version_id, options, pos, size, parents):
1304
"""Cache a version record in the history array and index cache.
1306
This is inlined into _load_data for performance. KEEP IN SYNC.
1307
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1310
# only want the _history index to reference the 1st index entry
1312
if version_id not in self._cache:
1313
index = len(self._history)
1314
self._history.append(version_id)
1316
index = self._cache[version_id][5]
1317
self._cache[version_id] = (version_id,
1324
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1325
create_parent_dir=False, delay_create=False, dir_mode=None):
1326
_KnitComponentFile.__init__(self, transport, filename, mode,
1327
file_mode=file_mode,
1328
create_parent_dir=create_parent_dir,
1331
# position in _history is the 'official' index for a revision
1332
# but the values may have come from a newer entry.
1333
# so - wc -l of a knit index is != the number of unique names
1337
fp = self._transport.get(self._filename)
1339
# _load_data may raise NoSuchFile if the target knit is
1341
_load_data(self, fp)
1345
if mode != 'w' or not create:
1348
self._need_to_create = True
1350
self._transport.put_bytes_non_atomic(
1351
self._filename, self.HEADER, mode=self._file_mode)
1353
def get_graph(self):
1354
"""Return a list of the node:parents lists from this knit index."""
1355
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1357
def get_ancestry(self, versions, topo_sorted=True):
1358
"""See VersionedFile.get_ancestry."""
1359
# get a graph of all the mentioned versions:
1361
pending = set(versions)
1364
version = pending.pop()
1367
parents = [p for p in cache[version][4] if p in cache]
1369
raise RevisionNotPresent(version, self._filename)
1370
# if not completed and not a ghost
1371
pending.update([p for p in parents if p not in graph])
1372
graph[version] = parents
1375
return topo_sort(graph.items())
1377
def get_ancestry_with_ghosts(self, versions):
1378
"""See VersionedFile.get_ancestry_with_ghosts."""
1379
# get a graph of all the mentioned versions:
1380
self.check_versions_present(versions)
1383
pending = set(versions)
1385
version = pending.pop()
1387
parents = cache[version][4]
1393
pending.update([p for p in parents if p not in graph])
1394
graph[version] = parents
1395
return topo_sort(graph.items())
1397
def iter_parents(self, version_ids):
1398
"""Iterate through the parents for many version ids.
1400
:param version_ids: An iterable yielding version_ids.
1401
:return: An iterator that yields (version_id, parents). Requested
1402
version_ids not present in the versioned file are simply skipped.
1403
The order is undefined, allowing for different optimisations in
1404
the underlying implementation.
1406
for version_id in version_ids:
1408
yield version_id, tuple(self.get_parents(version_id))
1412
def num_versions(self):
1413
return len(self._history)
1415
__len__ = num_versions
1417
def get_versions(self):
1418
"""Get all the versions in the file. not topologically sorted."""
1419
return self._history
1421
def _version_list_to_index(self, versions):
1424
for version in versions:
1425
if version in cache:
1426
# -- inlined lookup() --
1427
result_list.append(str(cache[version][5]))
1428
# -- end lookup () --
1430
result_list.append('.' + version)
1431
return ' '.join(result_list)
1433
def add_version(self, version_id, options, index_memo, parents):
1434
"""Add a version record to the index."""
1435
self.add_versions(((version_id, options, index_memo, parents),))
1437
def add_versions(self, versions, random_id=False):
1438
"""Add multiple versions to the index.
1440
:param versions: a list of tuples:
1441
(version_id, options, pos, size, parents).
1442
:param random_id: If True the ids being added were randomly generated
1443
and no check for existence will be performed.
1446
orig_history = self._history[:]
1447
orig_cache = self._cache.copy()
1450
for version_id, options, (index, pos, size), parents in versions:
1451
line = "\n%s %s %s %s %s :" % (version_id,
1455
self._version_list_to_index(parents))
1456
assert isinstance(line, str), \
1457
'content must be utf-8 encoded: %r' % (line,)
1459
self._cache_version(version_id, options, pos, size, parents)
1460
if not self._need_to_create:
1461
self._transport.append_bytes(self._filename, ''.join(lines))
1464
sio.write(self.HEADER)
1465
sio.writelines(lines)
1467
self._transport.put_file_non_atomic(self._filename, sio,
1468
create_parent_dir=self._create_parent_dir,
1469
mode=self._file_mode,
1470
dir_mode=self._dir_mode)
1471
self._need_to_create = False
1473
# If any problems happen, restore the original values and re-raise
1474
self._history = orig_history
1475
self._cache = orig_cache
1478
def has_version(self, version_id):
1479
"""True if the version is in the index."""
1480
return version_id in self._cache
1482
def get_position(self, version_id):
1483
"""Return details needed to access the version.
1485
.kndx indices do not support split-out data, so return None for the
1488
:return: a tuple (None, data position, size) to hand to the access
1489
logic to get the record.
1491
entry = self._cache[version_id]
1492
return None, entry[2], entry[3]
1494
def get_method(self, version_id):
1495
"""Return compression method of specified version."""
1497
options = self._cache[version_id][1]
1499
raise RevisionNotPresent(version_id, self._filename)
1500
if 'fulltext' in options:
1503
if 'line-delta' not in options:
1504
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1507
def get_options(self, version_id):
1508
"""Return a string represention options.
1512
return self._cache[version_id][1]
1514
def get_parents(self, version_id):
1515
"""Return parents of specified version ignoring ghosts."""
1516
return [parent for parent in self._cache[version_id][4]
1517
if parent in self._cache]
1519
def get_parents_with_ghosts(self, version_id):
1520
"""Return parents of specified version with ghosts."""
1521
return self._cache[version_id][4]
1523
def check_versions_present(self, version_ids):
1524
"""Check that all specified versions are present."""
1526
for version_id in version_ids:
1527
if version_id not in cache:
1528
raise RevisionNotPresent(version_id, self._filename)
1531
class KnitGraphIndex(object):
1532
"""A knit index that builds on GraphIndex."""
1534
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1535
"""Construct a KnitGraphIndex on a graph_index.
1537
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1538
:param deltas: Allow delta-compressed records.
1539
:param add_callback: If not None, allow additions to the index and call
1540
this callback with a list of added GraphIndex nodes:
1541
[(node, value, node_refs), ...]
1542
:param parents: If True, record knits parents, if not do not record
1545
self._graph_index = graph_index
1546
self._deltas = deltas
1547
self._add_callback = add_callback
1548
self._parents = parents
1549
if deltas and not parents:
1550
raise KnitCorrupt(self, "Cannot do delta compression without "
1553
def _get_entries(self, keys, check_present=False):
1554
"""Get the entries for keys.
1556
:param keys: An iterable of index keys, - 1-tuples.
1561
for node in self._graph_index.iter_entries(keys):
1563
found_keys.add(node[1])
1565
# adapt parentless index to the rest of the code.
1566
for node in self._graph_index.iter_entries(keys):
1567
yield node[0], node[1], node[2], ()
1568
found_keys.add(node[1])
1570
missing_keys = keys.difference(found_keys)
1572
raise RevisionNotPresent(missing_keys.pop(), self)
1574
def _present_keys(self, version_ids):
1576
node[1] for node in self._get_entries(version_ids)])
1578
def _parentless_ancestry(self, versions):
1579
"""Honour the get_ancestry API for parentless knit indices."""
1580
wanted_keys = self._version_ids_to_keys(versions)
1581
present_keys = self._present_keys(wanted_keys)
1582
missing = set(wanted_keys).difference(present_keys)
1584
raise RevisionNotPresent(missing.pop(), self)
1585
return list(self._keys_to_version_ids(present_keys))
1587
def get_ancestry(self, versions, topo_sorted=True):
1588
"""See VersionedFile.get_ancestry."""
1589
if not self._parents:
1590
return self._parentless_ancestry(versions)
1591
# XXX: This will do len(history) index calls - perhaps
1592
# it should be altered to be a index core feature?
1593
# get a graph of all the mentioned versions:
1596
versions = self._version_ids_to_keys(versions)
1597
pending = set(versions)
1599
# get all pending nodes
1600
this_iteration = pending
1601
new_nodes = self._get_entries(this_iteration)
1604
for (index, key, value, node_refs) in new_nodes:
1605
# dont ask for ghosties - otherwise
1606
# we we can end up looping with pending
1607
# being entirely ghosted.
1608
graph[key] = [parent for parent in node_refs[0]
1609
if parent not in ghosts]
1611
for parent in graph[key]:
1612
# dont examine known nodes again
1617
ghosts.update(this_iteration.difference(found))
1618
if versions.difference(graph):
1619
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1621
result_keys = topo_sort(graph.items())
1623
result_keys = graph.iterkeys()
1624
return [key[0] for key in result_keys]
1626
def get_ancestry_with_ghosts(self, versions):
1627
"""See VersionedFile.get_ancestry."""
1628
if not self._parents:
1629
return self._parentless_ancestry(versions)
1630
# XXX: This will do len(history) index calls - perhaps
1631
# it should be altered to be a index core feature?
1632
# get a graph of all the mentioned versions:
1634
versions = self._version_ids_to_keys(versions)
1635
pending = set(versions)
1637
# get all pending nodes
1638
this_iteration = pending
1639
new_nodes = self._get_entries(this_iteration)
1641
for (index, key, value, node_refs) in new_nodes:
1642
graph[key] = node_refs[0]
1644
for parent in graph[key]:
1645
# dont examine known nodes again
1649
missing_versions = this_iteration.difference(graph)
1650
missing_needed = versions.intersection(missing_versions)
1652
raise RevisionNotPresent(missing_needed.pop(), self)
1653
for missing_version in missing_versions:
1654
# add a key, no parents
1655
graph[missing_version] = []
1656
pending.discard(missing_version) # don't look for it
1657
result_keys = topo_sort(graph.items())
1658
return [key[0] for key in result_keys]
1660
def get_graph(self):
1661
"""Return a list of the node:parents lists from this knit index."""
1662
if not self._parents:
1663
return [(key, ()) for key in self.get_versions()]
1665
for index, key, value, refs in self._graph_index.iter_all_entries():
1666
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1669
def iter_parents(self, version_ids):
1670
"""Iterate through the parents for many version ids.
1672
:param version_ids: An iterable yielding version_ids.
1673
:return: An iterator that yields (version_id, parents). Requested
1674
version_ids not present in the versioned file are simply skipped.
1675
The order is undefined, allowing for different optimisations in
1676
the underlying implementation.
1679
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1681
present_parents = set()
1682
for node in all_nodes:
1683
all_parents.update(node[3][0])
1684
# any node we are querying must be present
1685
present_parents.add(node[1])
1686
unknown_parents = all_parents.difference(present_parents)
1687
present_parents.update(self._present_keys(unknown_parents))
1688
for node in all_nodes:
1690
for parent in node[3][0]:
1691
if parent in present_parents:
1692
parents.append(parent[0])
1693
yield node[1][0], tuple(parents)
1695
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1696
yield node[1][0], ()
1698
def num_versions(self):
1699
return len(list(self._graph_index.iter_all_entries()))
1701
__len__ = num_versions
1703
def get_versions(self):
1704
"""Get all the versions in the file. not topologically sorted."""
1705
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1707
def has_version(self, version_id):
1708
"""True if the version is in the index."""
1709
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1711
def _keys_to_version_ids(self, keys):
1712
return tuple(key[0] for key in keys)
1714
def get_position(self, version_id):
1715
"""Return details needed to access the version.
1717
:return: a tuple (index, data position, size) to hand to the access
1718
logic to get the record.
1720
node = self._get_node(version_id)
1721
bits = node[2][1:].split(' ')
1722
return node[0], int(bits[0]), int(bits[1])
1724
def get_method(self, version_id):
1725
"""Return compression method of specified version."""
1726
if not self._deltas:
1728
return self._parent_compression(self._get_node(version_id)[3][1])
1730
def _parent_compression(self, reference_list):
1731
# use the second reference list to decide if this is delta'd or not.
1732
if len(reference_list):
1737
def _get_node(self, version_id):
1739
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1741
raise RevisionNotPresent(version_id, self)
1743
def get_options(self, version_id):
1744
"""Return a string represention options.
1748
node = self._get_node(version_id)
1749
if not self._deltas:
1750
options = ['fulltext']
1752
options = [self._parent_compression(node[3][1])]
1753
if node[2][0] == 'N':
1754
options.append('no-eol')
1757
def get_parents(self, version_id):
1758
"""Return parents of specified version ignoring ghosts."""
1759
parents = list(self.iter_parents([version_id]))
1762
raise errors.RevisionNotPresent(version_id, self)
1763
return parents[0][1]
1765
def get_parents_with_ghosts(self, version_id):
1766
"""Return parents of specified version with ghosts."""
1767
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1768
check_present=True))
1769
if not self._parents:
1771
return self._keys_to_version_ids(nodes[0][3][0])
1773
def check_versions_present(self, version_ids):
1774
"""Check that all specified versions are present."""
1775
keys = self._version_ids_to_keys(version_ids)
1776
present = self._present_keys(keys)
1777
missing = keys.difference(present)
1779
raise RevisionNotPresent(missing.pop(), self)
1781
def add_version(self, version_id, options, access_memo, parents):
1782
"""Add a version record to the index."""
1783
return self.add_versions(((version_id, options, access_memo, parents),))
1785
def add_versions(self, versions, random_id=False):
1786
"""Add multiple versions to the index.
1788
This function does not insert data into the Immutable GraphIndex
1789
backing the KnitGraphIndex, instead it prepares data for insertion by
1790
the caller and checks that it is safe to insert then calls
1791
self._add_callback with the prepared GraphIndex nodes.
1793
:param versions: a list of tuples:
1794
(version_id, options, pos, size, parents).
1795
:param random_id: If True the ids being added were randomly generated
1796
and no check for existence will be performed.
1798
if not self._add_callback:
1799
raise errors.ReadOnlyError(self)
1800
# we hope there are no repositories with inconsistent parentage
1805
for (version_id, options, access_memo, parents) in versions:
1806
index, pos, size = access_memo
1807
key = (version_id, )
1808
parents = tuple((parent, ) for parent in parents)
1809
if 'no-eol' in options:
1813
value += "%d %d" % (pos, size)
1814
if not self._deltas:
1815
if 'line-delta' in options:
1816
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1819
if 'line-delta' in options:
1820
node_refs = (parents, (parents[0],))
1822
node_refs = (parents, ())
1824
node_refs = (parents, )
1827
raise KnitCorrupt(self, "attempt to add node with parents "
1828
"in parentless index.")
1830
keys[key] = (value, node_refs)
1832
present_nodes = self._get_entries(keys)
1833
for (index, key, value, node_refs) in present_nodes:
1834
if (value, node_refs) != keys[key]:
1835
raise KnitCorrupt(self, "inconsistent details in add_versions"
1836
": %s %s" % ((value, node_refs), keys[key]))
1840
for key, (value, node_refs) in keys.iteritems():
1841
result.append((key, value, node_refs))
1843
for key, (value, node_refs) in keys.iteritems():
1844
result.append((key, value))
1845
self._add_callback(result)
1847
def _version_ids_to_keys(self, version_ids):
1848
return set((version_id, ) for version_id in version_ids)
1851
class _KnitAccess(object):
1852
"""Access to knit records in a .knit file."""
1854
def __init__(self, transport, filename, _file_mode, _dir_mode,
1855
_need_to_create, _create_parent_dir):
1856
"""Create a _KnitAccess for accessing and inserting data.
1858
:param transport: The transport the .knit is located on.
1859
:param filename: The filename of the .knit.
1861
self._transport = transport
1862
self._filename = filename
1863
self._file_mode = _file_mode
1864
self._dir_mode = _dir_mode
1865
self._need_to_create = _need_to_create
1866
self._create_parent_dir = _create_parent_dir
1868
def add_raw_records(self, sizes, raw_data):
1869
"""Add raw knit bytes to a storage area.
1871
The data is spooled to whereever the access method is storing data.
1873
:param sizes: An iterable containing the size of each raw data segment.
1874
:param raw_data: A bytestring containing the data.
1875
:return: A list of memos to retrieve the record later. Each memo is a
1876
tuple - (index, pos, length), where the index field is always None
1877
for the .knit access method.
1879
assert type(raw_data) == str, \
1880
'data must be plain bytes was %s' % type(raw_data)
1881
if not self._need_to_create:
1882
base = self._transport.append_bytes(self._filename, raw_data)
1884
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1885
create_parent_dir=self._create_parent_dir,
1886
mode=self._file_mode,
1887
dir_mode=self._dir_mode)
1888
self._need_to_create = False
1892
result.append((None, base, size))
1897
"""IFF this data access has its own storage area, initialise it.
1901
self._transport.put_bytes_non_atomic(self._filename, '',
1902
mode=self._file_mode)
1904
def open_file(self):
1905
"""IFF this data access can be represented as a single file, open it.
1907
For knits that are not mapped to a single file on disk this will
1910
:return: None or a file handle.
1913
return self._transport.get(self._filename)
1918
def get_raw_records(self, memos_for_retrieval):
1919
"""Get the raw bytes for a records.
1921
:param memos_for_retrieval: An iterable containing the (index, pos,
1922
length) memo for retrieving the bytes. The .knit method ignores
1923
the index as there is always only a single file.
1924
:return: An iterator over the bytes of the records.
1926
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1927
for pos, data in self._transport.readv(self._filename, read_vector):
1931
class _PackAccess(object):
1932
"""Access to knit records via a collection of packs."""
1934
def __init__(self, index_to_packs, writer=None):
1935
"""Create a _PackAccess object.
1937
:param index_to_packs: A dict mapping index objects to the transport
1938
and file names for obtaining data.
1939
:param writer: A tuple (pack.ContainerWriter, write_index) which
1940
contains the pack to write, and the index that reads from it will
1944
self.container_writer = writer[0]
1945
self.write_index = writer[1]
1947
self.container_writer = None
1948
self.write_index = None
1949
self.indices = index_to_packs
1951
def add_raw_records(self, sizes, raw_data):
1952
"""Add raw knit bytes to a storage area.
1954
The data is spooled to the container writer in one bytes-record per
1957
:param sizes: An iterable containing the size of each raw data segment.
1958
:param raw_data: A bytestring containing the data.
1959
:return: A list of memos to retrieve the record later. Each memo is a
1960
tuple - (index, pos, length), where the index field is the
1961
write_index object supplied to the PackAccess object.
1963
assert type(raw_data) == str, \
1964
'data must be plain bytes was %s' % type(raw_data)
1968
p_offset, p_length = self.container_writer.add_bytes_record(
1969
raw_data[offset:offset+size], [])
1971
result.append((self.write_index, p_offset, p_length))
1975
"""Pack based knits do not get individually created."""
1977
def get_raw_records(self, memos_for_retrieval):
1978
"""Get the raw bytes for a records.
1980
:param memos_for_retrieval: An iterable containing the (index, pos,
1981
length) memo for retrieving the bytes. The Pack access method
1982
looks up the pack to use for a given record in its index_to_pack
1984
:return: An iterator over the bytes of the records.
1986
# first pass, group into same-index requests
1988
current_index = None
1989
for (index, offset, length) in memos_for_retrieval:
1990
if current_index == index:
1991
current_list.append((offset, length))
1993
if current_index is not None:
1994
request_lists.append((current_index, current_list))
1995
current_index = index
1996
current_list = [(offset, length)]
1997
# handle the last entry
1998
if current_index is not None:
1999
request_lists.append((current_index, current_list))
2000
for index, offsets in request_lists:
2001
transport, path = self.indices[index]
2002
reader = pack.make_readv_reader(transport, path, offsets)
2003
for names, read_func in reader.iter_records():
2004
yield read_func(None)
2006
def open_file(self):
2007
"""Pack based knits have no single file."""
2010
def set_writer(self, writer, index, (transport, packname)):
2011
"""Set a writer to use for adding data."""
2012
if index is not None:
2013
self.indices[index] = (transport, packname)
2014
self.container_writer = writer
2015
self.write_index = index
2018
class _KnitData(object):
2019
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2021
The KnitData class provides the logic for parsing and using knit records,
2022
making use of an access method for the low level read and write operations.
2025
def __init__(self, access):
2026
"""Create a KnitData object.
2028
:param access: The access method to use. Access methods such as
2029
_KnitAccess manage the insertion of raw records and the subsequent
2030
retrieval of the same.
2032
self._access = access
2033
self._checked = False
2034
# TODO: jam 20060713 conceptually, this could spill to disk
2035
# if the cached size gets larger than a certain amount
2036
# but it complicates the model a bit, so for now just use
2037
# a simple dictionary
2039
self._do_cache = False
2041
def enable_cache(self):
2042
"""Enable caching of reads."""
2043
self._do_cache = True
2045
def clear_cache(self):
2046
"""Clear the record cache."""
2047
self._do_cache = False
2050
def _open_file(self):
2051
return self._access.open_file()
2053
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2054
"""Convert version_id, digest, lines into a raw data block.
2056
:param dense_lines: The bytes of lines but in a denser form. For
2057
instance, if lines is a list of 1000 bytestrings each ending in \n,
2058
dense_lines may be a list with one line in it, containing all the
2059
1000's lines and their \n's. Using dense_lines if it is already
2060
known is a win because the string join to create bytes in this
2061
function spends less time resizing the final string.
2062
:return: (len, a StringIO instance with the raw data ready to read.)
2064
# Note: using a string copy here increases memory pressure with e.g.
2065
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2066
# when doing the initial commit of a mozilla tree. RBC 20070921
2067
bytes = ''.join(chain(
2068
["version %s %d %s\n" % (version_id,
2071
dense_lines or lines,
2072
["end %s\n" % version_id]))
2073
assert bytes.__class__ == str
2074
compressed_bytes = bytes_to_gzip(bytes)
2075
return len(compressed_bytes), compressed_bytes
2077
def add_raw_records(self, sizes, raw_data):
2078
"""Append a prepared record to the data file.
2080
:param sizes: An iterable containing the size of each raw data segment.
2081
:param raw_data: A bytestring containing the data.
2082
:return: a list of index data for the way the data was stored.
2083
See the access method add_raw_records documentation for more
2086
return self._access.add_raw_records(sizes, raw_data)
2088
def _parse_record_header(self, version_id, raw_data):
2089
"""Parse a record header for consistency.
2091
:return: the header and the decompressor stream.
2092
as (stream, header_record)
2094
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2096
rec = self._check_header(version_id, df.readline())
2097
except Exception, e:
2098
raise KnitCorrupt(self._access,
2099
"While reading {%s} got %s(%s)"
2100
% (version_id, e.__class__.__name__, str(e)))
2103
def _check_header(self, version_id, line):
2106
raise KnitCorrupt(self._access,
2107
'unexpected number of elements in record header')
2108
if rec[1] != version_id:
2109
raise KnitCorrupt(self._access,
2110
'unexpected version, wanted %r, got %r'
2111
% (version_id, rec[1]))
2114
def _parse_record(self, version_id, data):
2116
# 4168 calls in 2880 217 internal
2117
# 4168 calls to _parse_record_header in 2121
2118
# 4168 calls to readlines in 330
2119
df = GzipFile(mode='rb', fileobj=StringIO(data))
2122
record_contents = df.readlines()
2123
except Exception, e:
2124
raise KnitCorrupt(self._access,
2125
"While reading {%s} got %s(%s)"
2126
% (version_id, e.__class__.__name__, str(e)))
2127
header = record_contents.pop(0)
2128
rec = self._check_header(version_id, header)
2130
last_line = record_contents.pop()
2131
if len(record_contents) != int(rec[2]):
2132
raise KnitCorrupt(self._access,
2133
'incorrect number of lines %s != %s'
2135
% (len(record_contents), int(rec[2]),
2137
if last_line != 'end %s\n' % rec[1]:
2138
raise KnitCorrupt(self._access,
2139
'unexpected version end line %r, wanted %r'
2140
% (last_line, version_id))
2142
return record_contents, rec[3]
2144
def read_records_iter_raw(self, records):
2145
"""Read text records from data file and yield raw data.
2147
This unpacks enough of the text record to validate the id is
2148
as expected but thats all.
2150
# setup an iterator of the external records:
2151
# uses readv so nice and fast we hope.
2153
# grab the disk data needed.
2155
# Don't check _cache if it is empty
2156
needed_offsets = [index_memo for version_id, index_memo
2158
if version_id not in self._cache]
2160
needed_offsets = [index_memo for version_id, index_memo
2163
raw_records = self._access.get_raw_records(needed_offsets)
2165
for version_id, index_memo in records:
2166
if version_id in self._cache:
2167
# This data has already been validated
2168
data = self._cache[version_id]
2170
data = raw_records.next()
2172
self._cache[version_id] = data
2174
# validate the header
2175
df, rec = self._parse_record_header(version_id, data)
2177
yield version_id, data
2179
def read_records_iter(self, records):
2180
"""Read text records from data file and yield result.
2182
The result will be returned in whatever is the fastest to read.
2183
Not by the order requested. Also, multiple requests for the same
2184
record will only yield 1 response.
2185
:param records: A list of (version_id, pos, len) entries
2186
:return: Yields (version_id, contents, digest) in the order
2187
read, not the order requested
2193
# Skip records we have alread seen
2194
yielded_records = set()
2195
needed_records = set()
2196
for record in records:
2197
if record[0] in self._cache:
2198
if record[0] in yielded_records:
2200
yielded_records.add(record[0])
2201
data = self._cache[record[0]]
2202
content, digest = self._parse_record(record[0], data)
2203
yield (record[0], content, digest)
2205
needed_records.add(record)
2206
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2208
needed_records = sorted(set(records), key=operator.itemgetter(1))
2210
if not needed_records:
2213
# The transport optimizes the fetching as well
2214
# (ie, reads continuous ranges.)
2215
raw_data = self._access.get_raw_records(
2216
[index_memo for version_id, index_memo in needed_records])
2218
for (version_id, index_memo), data in \
2219
izip(iter(needed_records), raw_data):
2220
content, digest = self._parse_record(version_id, data)
2222
self._cache[version_id] = data
2223
yield version_id, content, digest
2225
def read_records(self, records):
2226
"""Read records into a dictionary."""
2228
for record_id, content, digest in \
2229
self.read_records_iter(records):
2230
components[record_id] = (content, digest)
2234
class InterKnit(InterVersionedFile):
2235
"""Optimised code paths for knit to knit operations."""
2237
_matching_file_from_factory = KnitVersionedFile
2238
_matching_file_to_factory = KnitVersionedFile
2241
def is_compatible(source, target):
2242
"""Be compatible with knits. """
2244
return (isinstance(source, KnitVersionedFile) and
2245
isinstance(target, KnitVersionedFile))
2246
except AttributeError:
2249
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2250
"""Copy texts to the target by extracting and adding them one by one.
2252
see join() for the parameter definitions.
2254
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2255
graph = self.source.get_graph(version_ids)
2256
order = topo_sort(graph.items())
2258
def size_of_content(content):
2259
return sum(len(line) for line in content.text())
2260
# Cache at most 10MB of parent texts
2261
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2262
compute_size=size_of_content)
2263
# TODO: jam 20071116 It would be nice to have a streaming interface to
2264
# get multiple texts from a source. The source could be smarter
2265
# about how it handled intermediate stages.
2266
# get_line_list() or make_mpdiffs() seem like a possibility, but
2267
# at the moment they extract all full texts into memory, which
2268
# causes us to store more than our 3x fulltext goal.
2269
# Repository.iter_files_bytes() may be another possibility
2270
to_process = [version for version in order
2271
if version not in self.target]
2272
total = len(to_process)
2273
pb = ui.ui_factory.nested_progress_bar()
2275
for index, version in enumerate(to_process):
2276
pb.update('Converting versioned data', index, total)
2277
sha1, num_bytes, parent_text = self.target.add_lines(version,
2278
self.source.get_parents(version),
2279
self.source.get_lines(version),
2280
parent_texts=parent_cache)
2281
parent_cache[version] = parent_text
2286
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2287
"""See InterVersionedFile.join."""
2288
assert isinstance(self.source, KnitVersionedFile)
2289
assert isinstance(self.target, KnitVersionedFile)
2291
# If the source and target are mismatched w.r.t. annotations vs
2292
# plain, the data needs to be converted accordingly
2293
if self.source.factory.annotated == self.target.factory.annotated:
2295
elif self.source.factory.annotated:
2296
converter = self._anno_to_plain_converter
2298
# We're converting from a plain to an annotated knit. Copy them
2299
# across by full texts.
2300
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2302
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2306
pb = ui.ui_factory.nested_progress_bar()
2308
version_ids = list(version_ids)
2309
if None in version_ids:
2310
version_ids.remove(None)
2312
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2313
this_versions = set(self.target._index.get_versions())
2314
# XXX: For efficiency we should not look at the whole index,
2315
# we only need to consider the referenced revisions - they
2316
# must all be present, or the method must be full-text.
2317
# TODO, RBC 20070919
2318
needed_versions = self.source_ancestry - this_versions
2320
if not needed_versions:
2322
full_list = topo_sort(self.source.get_graph())
2324
version_list = [i for i in full_list if (not self.target.has_version(i)
2325
and i in needed_versions)]
2329
copy_queue_records = []
2331
for version_id in version_list:
2332
options = self.source._index.get_options(version_id)
2333
parents = self.source._index.get_parents_with_ghosts(version_id)
2334
# check that its will be a consistent copy:
2335
for parent in parents:
2336
# if source has the parent, we must :
2337
# * already have it or
2338
# * have it scheduled already
2339
# otherwise we don't care
2340
assert (self.target.has_version(parent) or
2341
parent in copy_set or
2342
not self.source.has_version(parent))
2343
index_memo = self.source._index.get_position(version_id)
2344
copy_queue_records.append((version_id, index_memo))
2345
copy_queue.append((version_id, options, parents))
2346
copy_set.add(version_id)
2348
# data suck the join:
2350
total = len(version_list)
2353
for (version_id, raw_data), \
2354
(version_id2, options, parents) in \
2355
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2357
assert version_id == version_id2, 'logic error, inconsistent results'
2359
pb.update("Joining knit", count, total)
2361
size, raw_data = converter(raw_data, version_id, options,
2364
size = len(raw_data)
2365
raw_records.append((version_id, options, parents, size))
2366
raw_datum.append(raw_data)
2367
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2372
def _anno_to_plain_converter(self, raw_data, version_id, options,
2374
"""Convert annotated content to plain content."""
2375
data, digest = self.source._data._parse_record(version_id, raw_data)
2376
if 'fulltext' in options:
2377
content = self.source.factory.parse_fulltext(data, version_id)
2378
lines = self.target.factory.lower_fulltext(content)
2380
delta = self.source.factory.parse_line_delta(data, version_id,
2382
lines = self.target.factory.lower_line_delta(delta)
2383
return self.target._data._record_to_data(version_id, digest, lines)
2386
InterVersionedFile.register_optimiser(InterKnit)
2389
class WeaveToKnit(InterVersionedFile):
2390
"""Optimised code paths for weave to knit operations."""
2392
_matching_file_from_factory = bzrlib.weave.WeaveFile
2393
_matching_file_to_factory = KnitVersionedFile
2396
def is_compatible(source, target):
2397
"""Be compatible with weaves to knits."""
2399
return (isinstance(source, bzrlib.weave.Weave) and
2400
isinstance(target, KnitVersionedFile))
2401
except AttributeError:
2404
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2405
"""See InterVersionedFile.join."""
2406
assert isinstance(self.source, bzrlib.weave.Weave)
2407
assert isinstance(self.target, KnitVersionedFile)
2409
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2414
pb = ui.ui_factory.nested_progress_bar()
2416
version_ids = list(version_ids)
2418
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2419
this_versions = set(self.target._index.get_versions())
2420
needed_versions = self.source_ancestry - this_versions
2422
if not needed_versions:
2424
full_list = topo_sort(self.source.get_graph())
2426
version_list = [i for i in full_list if (not self.target.has_version(i)
2427
and i in needed_versions)]
2431
total = len(version_list)
2432
for version_id in version_list:
2433
pb.update("Converting to knit", count, total)
2434
parents = self.source.get_parents(version_id)
2435
# check that its will be a consistent copy:
2436
for parent in parents:
2437
# if source has the parent, we must already have it
2438
assert (self.target.has_version(parent))
2439
self.target.add_lines(
2440
version_id, parents, self.source.get_lines(version_id))
2447
InterVersionedFile.register_optimiser(WeaveToKnit)
2450
# Deprecated, use PatienceSequenceMatcher instead
2451
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2454
def annotate_knit(knit, revision_id):
2455
"""Annotate a knit with no cached annotations.
2457
This implementation is for knits with no cached annotations.
2458
It will work for knits with cached annotations, but this is not
2461
ancestry = knit.get_ancestry(revision_id)
2462
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2464
for candidate in ancestry:
2465
if candidate in annotations:
2467
parents = knit.get_parents(candidate)
2468
if len(parents) == 0:
2470
elif knit._index.get_method(candidate) != 'line-delta':
2473
parent, sha1, noeol, delta = knit.get_delta(candidate)
2474
blocks = KnitContent.get_line_delta_blocks(delta,
2475
fulltext[parents[0]], fulltext[candidate])
2476
annotations[candidate] = list(annotate.reannotate([annotations[p]
2477
for p in parents], fulltext[candidate], candidate, blocks))
2478
return iter(annotations[revision_id])
2482
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2484
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
