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
result_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
result_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"
636
# From here down to the return should really be logic in the returned
637
# callable -- in a class that adapts read_records_iter_raw to read
640
for (version_id, raw_data), \
641
(version_id2, options, _, parents) in \
642
izip(self._data.read_records_iter_raw(copy_queue_records),
643
result_version_list):
644
assert version_id == version_id2, \
645
'logic error, inconsistent results'
646
raw_datum.append(raw_data)
647
pseudo_file = StringIO(''.join(raw_datum))
650
return pseudo_file.read()
652
return pseudo_file.read(length)
653
return (self.get_format_signature(), result_version_list, read)
655
def _extract_blocks(self, version_id, source, target):
656
if self._index.get_method(version_id) != 'line-delta':
658
parent, sha1, noeol, delta = self.get_delta(version_id)
659
return KnitContent.get_line_delta_blocks(delta, source, target)
661
def get_delta(self, version_id):
662
"""Get a delta for constructing version from some other version."""
663
self.check_not_reserved_id(version_id)
664
parents = self.get_parents(version_id)
669
index_memo = self._index.get_position(version_id)
670
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
671
noeol = 'no-eol' in self._index.get_options(version_id)
672
if 'fulltext' == self._index.get_method(version_id):
673
new_content = self.factory.parse_fulltext(data, version_id)
674
if parent is not None:
675
reference_content = self._get_content(parent)
676
old_texts = reference_content.text()
679
new_texts = new_content.text()
680
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
682
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
684
delta = self.factory.parse_line_delta(data, version_id)
685
return parent, sha1, noeol, delta
687
def get_format_signature(self):
688
"""See VersionedFile.get_format_signature()."""
689
if self.factory.annotated:
690
annotated_part = "annotated"
692
annotated_part = "plain"
693
return "knit-%s" % (annotated_part,)
695
def get_graph_with_ghosts(self):
696
"""See VersionedFile.get_graph_with_ghosts()."""
697
graph_items = self._index.get_graph()
698
return dict(graph_items)
700
def get_sha1(self, version_id):
701
return self.get_sha1s([version_id])[0]
703
def get_sha1s(self, version_ids):
704
"""See VersionedFile.get_sha1()."""
705
record_map = self._get_record_map(version_ids)
706
# record entry 2 is the 'digest'.
707
return [record_map[v][2] for v in version_ids]
711
"""See VersionedFile.get_suffixes()."""
712
return [DATA_SUFFIX, INDEX_SUFFIX]
714
def has_ghost(self, version_id):
715
"""True if there is a ghost reference in the file to version_id."""
717
if self.has_version(version_id):
719
# optimisable if needed by memoising the _ghosts set.
720
items = self._index.get_graph()
721
for node, parents in items:
722
for parent in parents:
723
if parent not in self._index._cache:
724
if parent == version_id:
728
def insert_data_stream(self, (format, data_list, reader_callable)):
729
"""Insert knit records from a data stream into this knit.
731
If a version in the stream is already present in this knit, it will not
732
be inserted a second time. It will be checked for consistency with the
733
stored version however, and may cause a KnitCorrupt error to be raised
734
if the data in the stream disagrees with the already stored data.
736
:seealso: get_data_stream
738
if format != self.get_format_signature():
739
trace.mutter('incompatible format signature inserting to %r', self)
740
raise KnitDataStreamIncompatible(
741
format, self.get_format_signature())
743
for version_id, options, length, parents in data_list:
744
if self.has_version(version_id):
745
# First check: the list of parents.
746
my_parents = self.get_parents_with_ghosts(version_id)
747
if my_parents != parents:
748
# XXX: KnitCorrupt is not quite the right exception here.
751
'parents list %r from data stream does not match '
752
'already recorded parents %r for %s'
753
% (parents, my_parents, version_id))
755
# Also check the SHA-1 of the fulltext this content will
757
raw_data = reader_callable(length)
758
my_fulltext_sha1 = self.get_sha1(version_id)
759
df, rec = self._data._parse_record_header(version_id, raw_data)
760
stream_fulltext_sha1 = rec[3]
761
if my_fulltext_sha1 != stream_fulltext_sha1:
762
# Actually, we don't know if it's this knit that's corrupt,
763
# or the data stream we're trying to insert.
765
self.filename, 'sha-1 does not match %s' % version_id)
767
if 'line-delta' in options:
768
# Make sure that this knit record is actually useful: a
769
# line-delta is no use unless we have its parent.
770
# Fetching from a broken repository with this problem
771
# shouldn't break the target repository.
772
if not self._index.has_version(parents[0]):
775
'line-delta from stream references '
776
'missing parent %s' % parents[0])
777
self._add_raw_records(
778
[(version_id, options, parents, length)],
779
reader_callable(length))
782
"""See VersionedFile.versions."""
783
if 'evil' in debug.debug_flags:
784
trace.mutter_callsite(2, "versions scales with size of history")
785
return self._index.get_versions()
787
def has_version(self, version_id):
788
"""See VersionedFile.has_version."""
789
if 'evil' in debug.debug_flags:
790
trace.mutter_callsite(2, "has_version is a LBYL scenario")
791
return self._index.has_version(version_id)
793
__contains__ = has_version
795
def _merge_annotations(self, content, parents, parent_texts={},
796
delta=None, annotated=None,
797
left_matching_blocks=None):
798
"""Merge annotations for content. This is done by comparing
799
the annotations based on changed to the text.
801
if left_matching_blocks is not None:
802
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
806
for parent_id in parents:
807
merge_content = self._get_content(parent_id, parent_texts)
808
if (parent_id == parents[0] and delta_seq is not None):
811
seq = patiencediff.PatienceSequenceMatcher(
812
None, merge_content.text(), content.text())
813
for i, j, n in seq.get_matching_blocks():
816
# this appears to copy (origin, text) pairs across to the
817
# new content for any line that matches the last-checked
819
content._lines[j:j+n] = merge_content._lines[i:i+n]
821
if delta_seq is None:
822
reference_content = self._get_content(parents[0], parent_texts)
823
new_texts = content.text()
824
old_texts = reference_content.text()
825
delta_seq = patiencediff.PatienceSequenceMatcher(
826
None, old_texts, new_texts)
827
return self._make_line_delta(delta_seq, content)
829
def _make_line_delta(self, delta_seq, new_content):
830
"""Generate a line delta from delta_seq and new_content."""
832
for op in delta_seq.get_opcodes():
835
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
838
def _get_components_positions(self, version_ids):
839
"""Produce a map of position data for the components of versions.
841
This data is intended to be used for retrieving the knit records.
843
A dict of version_id to (method, data_pos, data_size, next) is
845
method is the way referenced data should be applied.
846
data_pos is the position of the data in the knit.
847
data_size is the size of the data in the knit.
848
next is the build-parent of the version, or None for fulltexts.
851
for version_id in version_ids:
854
while cursor is not None and cursor not in component_data:
855
method = self._index.get_method(cursor)
856
if method == 'fulltext':
859
next = self.get_parents_with_ghosts(cursor)[0]
860
index_memo = self._index.get_position(cursor)
861
component_data[cursor] = (method, index_memo, next)
863
return component_data
865
def _get_content(self, version_id, parent_texts={}):
866
"""Returns a content object that makes up the specified
868
cached_version = parent_texts.get(version_id, None)
869
if cached_version is not None:
870
if not self.has_version(version_id):
871
raise RevisionNotPresent(version_id, self.filename)
872
return cached_version
874
text_map, contents_map = self._get_content_maps([version_id])
875
return contents_map[version_id]
877
def _check_versions_present(self, version_ids):
878
"""Check that all specified versions are present."""
879
self._index.check_versions_present(version_ids)
881
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
882
nostore_sha, random_id, check_content):
883
"""See VersionedFile.add_lines_with_ghosts()."""
884
self._check_add(version_id, lines, random_id, check_content)
885
return self._add(version_id, lines, parents, self.delta,
886
parent_texts, None, nostore_sha, random_id)
888
def _add_lines(self, version_id, parents, lines, parent_texts,
889
left_matching_blocks, nostore_sha, random_id, check_content):
890
"""See VersionedFile.add_lines."""
891
self._check_add(version_id, lines, random_id, check_content)
892
self._check_versions_present(parents)
893
return self._add(version_id, lines[:], parents, self.delta,
894
parent_texts, left_matching_blocks, nostore_sha, random_id)
896
def _check_add(self, version_id, lines, random_id, check_content):
897
"""check that version_id and lines are safe to add."""
898
if contains_whitespace(version_id):
899
raise InvalidRevisionId(version_id, self.filename)
900
self.check_not_reserved_id(version_id)
901
# Technically this could be avoided if we are happy to allow duplicate
902
# id insertion when other things than bzr core insert texts, but it
903
# seems useful for folk using the knit api directly to have some safety
904
# blanket that we can disable.
905
if not random_id and self.has_version(version_id):
906
raise RevisionAlreadyPresent(version_id, self.filename)
908
self._check_lines_not_unicode(lines)
909
self._check_lines_are_lines(lines)
911
def _add(self, version_id, lines, parents, delta, parent_texts,
912
left_matching_blocks, nostore_sha, random_id):
913
"""Add a set of lines on top of version specified by parents.
915
If delta is true, compress the text as a line-delta against
918
Any versions not present will be converted into ghosts.
920
# first thing, if the content is something we don't need to store, find
922
line_bytes = ''.join(lines)
923
digest = sha_string(line_bytes)
924
if nostore_sha == digest:
925
raise errors.ExistingContent
928
if parent_texts is None:
930
for parent in parents:
931
if self.has_version(parent):
932
present_parents.append(parent)
934
# can only compress against the left most present parent.
936
(len(present_parents) == 0 or
937
present_parents[0] != parents[0])):
940
text_length = len(line_bytes)
943
if lines[-1][-1] != '\n':
944
# copy the contents of lines.
946
options.append('no-eol')
947
lines[-1] = lines[-1] + '\n'
951
# To speed the extract of texts the delta chain is limited
952
# to a fixed number of deltas. This should minimize both
953
# I/O and the time spend applying deltas.
954
delta = self._check_should_delta(present_parents)
956
assert isinstance(version_id, str)
957
content = self.factory.make(lines, version_id)
958
if delta or (self.factory.annotated and len(present_parents) > 0):
959
# Merge annotations from parent texts if needed.
960
delta_hunks = self._merge_annotations(content, present_parents,
961
parent_texts, delta, self.factory.annotated,
962
left_matching_blocks)
965
options.append('line-delta')
966
store_lines = self.factory.lower_line_delta(delta_hunks)
967
size, bytes = self._data._record_to_data(version_id, digest,
970
options.append('fulltext')
971
# isinstance is slower and we have no hierarchy.
972
if self.factory.__class__ == KnitPlainFactory:
973
# Use the already joined bytes saving iteration time in
975
size, bytes = self._data._record_to_data(version_id, digest,
978
# get mixed annotation + content and feed it into the
980
store_lines = self.factory.lower_fulltext(content)
981
size, bytes = self._data._record_to_data(version_id, digest,
984
access_memo = self._data.add_raw_records([size], bytes)[0]
985
self._index.add_versions(
986
((version_id, options, access_memo, parents),),
988
return digest, text_length, content
990
def check(self, progress_bar=None):
991
"""See VersionedFile.check()."""
993
def _clone_text(self, new_version_id, old_version_id, parents):
994
"""See VersionedFile.clone_text()."""
995
# FIXME RBC 20060228 make fast by only inserting an index with null
997
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
999
def get_lines(self, version_id):
1000
"""See VersionedFile.get_lines()."""
1001
return self.get_line_list([version_id])[0]
1003
def _get_record_map(self, version_ids):
1004
"""Produce a dictionary of knit records.
1006
The keys are version_ids, the values are tuples of (method, content,
1008
method is the way the content should be applied.
1009
content is a KnitContent object.
1010
digest is the SHA1 digest of this version id after all steps are done
1011
next is the build-parent of the version, i.e. the leftmost ancestor.
1012
If the method is fulltext, next will be None.
1014
position_map = self._get_components_positions(version_ids)
1015
# c = component_id, m = method, i_m = index_memo, n = next
1016
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1018
for component_id, content, digest in \
1019
self._data.read_records_iter(records):
1020
method, index_memo, next = position_map[component_id]
1021
record_map[component_id] = method, content, digest, next
1025
def get_text(self, version_id):
1026
"""See VersionedFile.get_text"""
1027
return self.get_texts([version_id])[0]
1029
def get_texts(self, version_ids):
1030
return [''.join(l) for l in self.get_line_list(version_ids)]
1032
def get_line_list(self, version_ids):
1033
"""Return the texts of listed versions as a list of strings."""
1034
for version_id in version_ids:
1035
self.check_not_reserved_id(version_id)
1036
text_map, content_map = self._get_content_maps(version_ids)
1037
return [text_map[v] for v in version_ids]
1039
_get_lf_split_line_list = get_line_list
1041
def _get_content_maps(self, version_ids):
1042
"""Produce maps of text and KnitContents
1044
:return: (text_map, content_map) where text_map contains the texts for
1045
the requested versions and content_map contains the KnitContents.
1046
Both dicts take version_ids as their keys.
1048
# FUTURE: This function could be improved for the 'extract many' case
1049
# by tracking each component and only doing the copy when the number of
1050
# children than need to apply delta's to it is > 1 or it is part of the
1052
version_ids = list(version_ids)
1053
multiple_versions = len(version_ids) != 1
1054
record_map = self._get_record_map(version_ids)
1059
for version_id in version_ids:
1062
while cursor is not None:
1063
method, data, digest, next = record_map[cursor]
1064
components.append((cursor, method, data, digest))
1065
if cursor in content_map:
1070
for component_id, method, data, digest in reversed(components):
1071
if component_id in content_map:
1072
content = content_map[component_id]
1074
if method == 'fulltext':
1075
assert content is None
1076
content = self.factory.parse_fulltext(data, version_id)
1077
elif method == 'line-delta':
1078
delta = self.factory.parse_line_delta(data, version_id)
1079
if multiple_versions:
1080
# only doing this when we want multiple versions
1081
# output avoids list copies - which reference and
1082
# dereference many strings.
1083
content = content.copy()
1084
content.apply_delta(delta, version_id)
1085
if multiple_versions:
1086
content_map[component_id] = content
1088
if 'no-eol' in self._index.get_options(version_id):
1089
if multiple_versions:
1090
content = content.copy()
1091
content.strip_last_line_newline()
1092
final_content[version_id] = content
1094
# digest here is the digest from the last applied component.
1095
text = content.text()
1096
actual_sha = sha_strings(text)
1097
if actual_sha != digest:
1098
raise KnitCorrupt(self.filename,
1100
'\n of reconstructed text does not match'
1102
'\n for version %s' %
1103
(actual_sha, digest, version_id))
1104
text_map[version_id] = text
1105
return text_map, final_content
1107
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1109
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1110
if version_ids is None:
1111
version_ids = self.versions()
1113
pb = progress.DummyProgress()
1114
# we don't care about inclusions, the caller cares.
1115
# but we need to setup a list of records to visit.
1116
# we need version_id, position, length
1117
version_id_records = []
1118
requested_versions = set(version_ids)
1119
# filter for available versions
1120
for version_id in requested_versions:
1121
if not self.has_version(version_id):
1122
raise RevisionNotPresent(version_id, self.filename)
1123
# get a in-component-order queue:
1124
for version_id in self.versions():
1125
if version_id in requested_versions:
1126
index_memo = self._index.get_position(version_id)
1127
version_id_records.append((version_id, index_memo))
1129
total = len(version_id_records)
1130
for version_idx, (version_id, data, sha_value) in \
1131
enumerate(self._data.read_records_iter(version_id_records)):
1132
pb.update('Walking content.', version_idx, total)
1133
method = self._index.get_method(version_id)
1135
assert method in ('fulltext', 'line-delta')
1136
if method == 'fulltext':
1137
line_iterator = self.factory.get_fulltext_content(data)
1139
line_iterator = self.factory.get_linedelta_content(data)
1140
# XXX: It might be more efficient to yield (version_id,
1141
# line_iterator) in the future. However for now, this is a simpler
1142
# change to integrate into the rest of the codebase. RBC 20071110
1143
for line in line_iterator:
1144
yield line, version_id
1146
pb.update('Walking content.', total, total)
1148
def iter_parents(self, version_ids):
1149
"""Iterate through the parents for many version ids.
1151
:param version_ids: An iterable yielding version_ids.
1152
:return: An iterator that yields (version_id, parents). Requested
1153
version_ids not present in the versioned file are simply skipped.
1154
The order is undefined, allowing for different optimisations in
1155
the underlying implementation.
1157
return self._index.iter_parents(version_ids)
1159
def num_versions(self):
1160
"""See VersionedFile.num_versions()."""
1161
return self._index.num_versions()
1163
__len__ = num_versions
1165
def annotate_iter(self, version_id):
1166
"""See VersionedFile.annotate_iter."""
1167
return self.factory.annotate_iter(self, version_id)
1169
def get_parents(self, version_id):
1170
"""See VersionedFile.get_parents."""
1173
# 52554 calls in 1264 872 internal down from 3674
1175
return self._index.get_parents(version_id)
1177
raise RevisionNotPresent(version_id, self.filename)
1179
def get_parents_with_ghosts(self, version_id):
1180
"""See VersionedFile.get_parents."""
1182
return self._index.get_parents_with_ghosts(version_id)
1184
raise RevisionNotPresent(version_id, self.filename)
1186
def get_ancestry(self, versions, topo_sorted=True):
1187
"""See VersionedFile.get_ancestry."""
1188
if isinstance(versions, basestring):
1189
versions = [versions]
1192
return self._index.get_ancestry(versions, topo_sorted)
1194
def get_ancestry_with_ghosts(self, versions):
1195
"""See VersionedFile.get_ancestry_with_ghosts."""
1196
if isinstance(versions, basestring):
1197
versions = [versions]
1200
return self._index.get_ancestry_with_ghosts(versions)
1202
def plan_merge(self, ver_a, ver_b):
1203
"""See VersionedFile.plan_merge."""
1204
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1205
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1206
annotated_a = self.annotate(ver_a)
1207
annotated_b = self.annotate(ver_b)
1208
return merge._plan_annotate_merge(annotated_a, annotated_b,
1209
ancestors_a, ancestors_b)
1212
class _KnitComponentFile(object):
1213
"""One of the files used to implement a knit database"""
1215
def __init__(self, transport, filename, mode, file_mode=None,
1216
create_parent_dir=False, dir_mode=None):
1217
self._transport = transport
1218
self._filename = filename
1220
self._file_mode = file_mode
1221
self._dir_mode = dir_mode
1222
self._create_parent_dir = create_parent_dir
1223
self._need_to_create = False
1225
def _full_path(self):
1226
"""Return the full path to this file."""
1227
return self._transport.base + self._filename
1229
def check_header(self, fp):
1230
line = fp.readline()
1232
# An empty file can actually be treated as though the file doesn't
1234
raise errors.NoSuchFile(self._full_path())
1235
if line != self.HEADER:
1236
raise KnitHeaderError(badline=line,
1237
filename=self._transport.abspath(self._filename))
1240
return '%s(%s)' % (self.__class__.__name__, self._filename)
1243
class _KnitIndex(_KnitComponentFile):
1244
"""Manages knit index file.
1246
The index is already kept in memory and read on startup, to enable
1247
fast lookups of revision information. The cursor of the index
1248
file is always pointing to the end, making it easy to append
1251
_cache is a cache for fast mapping from version id to a Index
1254
_history is a cache for fast mapping from indexes to version ids.
1256
The index data format is dictionary compressed when it comes to
1257
parent references; a index entry may only have parents that with a
1258
lover index number. As a result, the index is topological sorted.
1260
Duplicate entries may be written to the index for a single version id
1261
if this is done then the latter one completely replaces the former:
1262
this allows updates to correct version and parent information.
1263
Note that the two entries may share the delta, and that successive
1264
annotations and references MUST point to the first entry.
1266
The index file on disc contains a header, followed by one line per knit
1267
record. The same revision can be present in an index file more than once.
1268
The first occurrence gets assigned a sequence number starting from 0.
1270
The format of a single line is
1271
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1272
REVISION_ID is a utf8-encoded revision id
1273
FLAGS is a comma separated list of flags about the record. Values include
1274
no-eol, line-delta, fulltext.
1275
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1276
that the the compressed data starts at.
1277
LENGTH is the ascii representation of the length of the data file.
1278
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1280
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1281
revision id already in the knit that is a parent of REVISION_ID.
1282
The ' :' marker is the end of record marker.
1285
when a write is interrupted to the index file, it will result in a line
1286
that does not end in ' :'. If the ' :' is not present at the end of a line,
1287
or at the end of the file, then the record that is missing it will be
1288
ignored by the parser.
1290
When writing new records to the index file, the data is preceded by '\n'
1291
to ensure that records always start on new lines even if the last write was
1292
interrupted. As a result its normal for the last line in the index to be
1293
missing a trailing newline. One can be added with no harmful effects.
1296
HEADER = "# bzr knit index 8\n"
1298
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1299
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1301
def _cache_version(self, version_id, options, pos, size, parents):
1302
"""Cache a version record in the history array and index cache.
1304
This is inlined into _load_data for performance. KEEP IN SYNC.
1305
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1308
# only want the _history index to reference the 1st index entry
1310
if version_id not in self._cache:
1311
index = len(self._history)
1312
self._history.append(version_id)
1314
index = self._cache[version_id][5]
1315
self._cache[version_id] = (version_id,
1322
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1323
create_parent_dir=False, delay_create=False, dir_mode=None):
1324
_KnitComponentFile.__init__(self, transport, filename, mode,
1325
file_mode=file_mode,
1326
create_parent_dir=create_parent_dir,
1329
# position in _history is the 'official' index for a revision
1330
# but the values may have come from a newer entry.
1331
# so - wc -l of a knit index is != the number of unique names
1335
fp = self._transport.get(self._filename)
1337
# _load_data may raise NoSuchFile if the target knit is
1339
_load_data(self, fp)
1343
if mode != 'w' or not create:
1346
self._need_to_create = True
1348
self._transport.put_bytes_non_atomic(
1349
self._filename, self.HEADER, mode=self._file_mode)
1351
def get_graph(self):
1352
"""Return a list of the node:parents lists from this knit index."""
1353
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1355
def get_ancestry(self, versions, topo_sorted=True):
1356
"""See VersionedFile.get_ancestry."""
1357
# get a graph of all the mentioned versions:
1359
pending = set(versions)
1362
version = pending.pop()
1365
parents = [p for p in cache[version][4] if p in cache]
1367
raise RevisionNotPresent(version, self._filename)
1368
# if not completed and not a ghost
1369
pending.update([p for p in parents if p not in graph])
1370
graph[version] = parents
1373
return topo_sort(graph.items())
1375
def get_ancestry_with_ghosts(self, versions):
1376
"""See VersionedFile.get_ancestry_with_ghosts."""
1377
# get a graph of all the mentioned versions:
1378
self.check_versions_present(versions)
1381
pending = set(versions)
1383
version = pending.pop()
1385
parents = cache[version][4]
1391
pending.update([p for p in parents if p not in graph])
1392
graph[version] = parents
1393
return topo_sort(graph.items())
1395
def iter_parents(self, version_ids):
1396
"""Iterate through the parents for many version ids.
1398
:param version_ids: An iterable yielding version_ids.
1399
:return: An iterator that yields (version_id, parents). Requested
1400
version_ids not present in the versioned file are simply skipped.
1401
The order is undefined, allowing for different optimisations in
1402
the underlying implementation.
1404
for version_id in version_ids:
1406
yield version_id, tuple(self.get_parents(version_id))
1410
def num_versions(self):
1411
return len(self._history)
1413
__len__ = num_versions
1415
def get_versions(self):
1416
"""Get all the versions in the file. not topologically sorted."""
1417
return self._history
1419
def _version_list_to_index(self, versions):
1422
for version in versions:
1423
if version in cache:
1424
# -- inlined lookup() --
1425
result_list.append(str(cache[version][5]))
1426
# -- end lookup () --
1428
result_list.append('.' + version)
1429
return ' '.join(result_list)
1431
def add_version(self, version_id, options, index_memo, parents):
1432
"""Add a version record to the index."""
1433
self.add_versions(((version_id, options, index_memo, parents),))
1435
def add_versions(self, versions, random_id=False):
1436
"""Add multiple versions to the index.
1438
:param versions: a list of tuples:
1439
(version_id, options, pos, size, parents).
1440
:param random_id: If True the ids being added were randomly generated
1441
and no check for existence will be performed.
1444
orig_history = self._history[:]
1445
orig_cache = self._cache.copy()
1448
for version_id, options, (index, pos, size), parents in versions:
1449
line = "\n%s %s %s %s %s :" % (version_id,
1453
self._version_list_to_index(parents))
1454
assert isinstance(line, str), \
1455
'content must be utf-8 encoded: %r' % (line,)
1457
self._cache_version(version_id, options, pos, size, parents)
1458
if not self._need_to_create:
1459
self._transport.append_bytes(self._filename, ''.join(lines))
1462
sio.write(self.HEADER)
1463
sio.writelines(lines)
1465
self._transport.put_file_non_atomic(self._filename, sio,
1466
create_parent_dir=self._create_parent_dir,
1467
mode=self._file_mode,
1468
dir_mode=self._dir_mode)
1469
self._need_to_create = False
1471
# If any problems happen, restore the original values and re-raise
1472
self._history = orig_history
1473
self._cache = orig_cache
1476
def has_version(self, version_id):
1477
"""True if the version is in the index."""
1478
return version_id in self._cache
1480
def get_position(self, version_id):
1481
"""Return details needed to access the version.
1483
.kndx indices do not support split-out data, so return None for the
1486
:return: a tuple (None, data position, size) to hand to the access
1487
logic to get the record.
1489
entry = self._cache[version_id]
1490
return None, entry[2], entry[3]
1492
def get_method(self, version_id):
1493
"""Return compression method of specified version."""
1495
options = self._cache[version_id][1]
1497
raise RevisionNotPresent(version_id, self._filename)
1498
if 'fulltext' in options:
1501
if 'line-delta' not in options:
1502
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1505
def get_options(self, version_id):
1506
"""Return a string represention options.
1510
return self._cache[version_id][1]
1512
def get_parents(self, version_id):
1513
"""Return parents of specified version ignoring ghosts."""
1514
return [parent for parent in self._cache[version_id][4]
1515
if parent in self._cache]
1517
def get_parents_with_ghosts(self, version_id):
1518
"""Return parents of specified version with ghosts."""
1519
return self._cache[version_id][4]
1521
def check_versions_present(self, version_ids):
1522
"""Check that all specified versions are present."""
1524
for version_id in version_ids:
1525
if version_id not in cache:
1526
raise RevisionNotPresent(version_id, self._filename)
1529
class KnitGraphIndex(object):
1530
"""A knit index that builds on GraphIndex."""
1532
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1533
"""Construct a KnitGraphIndex on a graph_index.
1535
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1536
:param deltas: Allow delta-compressed records.
1537
:param add_callback: If not None, allow additions to the index and call
1538
this callback with a list of added GraphIndex nodes:
1539
[(node, value, node_refs), ...]
1540
:param parents: If True, record knits parents, if not do not record
1543
self._graph_index = graph_index
1544
self._deltas = deltas
1545
self._add_callback = add_callback
1546
self._parents = parents
1547
if deltas and not parents:
1548
raise KnitCorrupt(self, "Cannot do delta compression without "
1551
def _get_entries(self, keys, check_present=False):
1552
"""Get the entries for keys.
1554
:param keys: An iterable of index keys, - 1-tuples.
1559
for node in self._graph_index.iter_entries(keys):
1561
found_keys.add(node[1])
1563
# adapt parentless index to the rest of the code.
1564
for node in self._graph_index.iter_entries(keys):
1565
yield node[0], node[1], node[2], ()
1566
found_keys.add(node[1])
1568
missing_keys = keys.difference(found_keys)
1570
raise RevisionNotPresent(missing_keys.pop(), self)
1572
def _present_keys(self, version_ids):
1574
node[1] for node in self._get_entries(version_ids)])
1576
def _parentless_ancestry(self, versions):
1577
"""Honour the get_ancestry API for parentless knit indices."""
1578
wanted_keys = self._version_ids_to_keys(versions)
1579
present_keys = self._present_keys(wanted_keys)
1580
missing = set(wanted_keys).difference(present_keys)
1582
raise RevisionNotPresent(missing.pop(), self)
1583
return list(self._keys_to_version_ids(present_keys))
1585
def get_ancestry(self, versions, topo_sorted=True):
1586
"""See VersionedFile.get_ancestry."""
1587
if not self._parents:
1588
return self._parentless_ancestry(versions)
1589
# XXX: This will do len(history) index calls - perhaps
1590
# it should be altered to be a index core feature?
1591
# get a graph of all the mentioned versions:
1594
versions = self._version_ids_to_keys(versions)
1595
pending = set(versions)
1597
# get all pending nodes
1598
this_iteration = pending
1599
new_nodes = self._get_entries(this_iteration)
1602
for (index, key, value, node_refs) in new_nodes:
1603
# dont ask for ghosties - otherwise
1604
# we we can end up looping with pending
1605
# being entirely ghosted.
1606
graph[key] = [parent for parent in node_refs[0]
1607
if parent not in ghosts]
1609
for parent in graph[key]:
1610
# dont examine known nodes again
1615
ghosts.update(this_iteration.difference(found))
1616
if versions.difference(graph):
1617
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1619
result_keys = topo_sort(graph.items())
1621
result_keys = graph.iterkeys()
1622
return [key[0] for key in result_keys]
1624
def get_ancestry_with_ghosts(self, versions):
1625
"""See VersionedFile.get_ancestry."""
1626
if not self._parents:
1627
return self._parentless_ancestry(versions)
1628
# XXX: This will do len(history) index calls - perhaps
1629
# it should be altered to be a index core feature?
1630
# get a graph of all the mentioned versions:
1632
versions = self._version_ids_to_keys(versions)
1633
pending = set(versions)
1635
# get all pending nodes
1636
this_iteration = pending
1637
new_nodes = self._get_entries(this_iteration)
1639
for (index, key, value, node_refs) in new_nodes:
1640
graph[key] = node_refs[0]
1642
for parent in graph[key]:
1643
# dont examine known nodes again
1647
missing_versions = this_iteration.difference(graph)
1648
missing_needed = versions.intersection(missing_versions)
1650
raise RevisionNotPresent(missing_needed.pop(), self)
1651
for missing_version in missing_versions:
1652
# add a key, no parents
1653
graph[missing_version] = []
1654
pending.discard(missing_version) # don't look for it
1655
result_keys = topo_sort(graph.items())
1656
return [key[0] for key in result_keys]
1658
def get_graph(self):
1659
"""Return a list of the node:parents lists from this knit index."""
1660
if not self._parents:
1661
return [(key, ()) for key in self.get_versions()]
1663
for index, key, value, refs in self._graph_index.iter_all_entries():
1664
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1667
def iter_parents(self, version_ids):
1668
"""Iterate through the parents for many version ids.
1670
:param version_ids: An iterable yielding version_ids.
1671
:return: An iterator that yields (version_id, parents). Requested
1672
version_ids not present in the versioned file are simply skipped.
1673
The order is undefined, allowing for different optimisations in
1674
the underlying implementation.
1677
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1679
present_parents = set()
1680
for node in all_nodes:
1681
all_parents.update(node[3][0])
1682
# any node we are querying must be present
1683
present_parents.add(node[1])
1684
unknown_parents = all_parents.difference(present_parents)
1685
present_parents.update(self._present_keys(unknown_parents))
1686
for node in all_nodes:
1688
for parent in node[3][0]:
1689
if parent in present_parents:
1690
parents.append(parent[0])
1691
yield node[1][0], tuple(parents)
1693
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1694
yield node[1][0], ()
1696
def num_versions(self):
1697
return len(list(self._graph_index.iter_all_entries()))
1699
__len__ = num_versions
1701
def get_versions(self):
1702
"""Get all the versions in the file. not topologically sorted."""
1703
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1705
def has_version(self, version_id):
1706
"""True if the version is in the index."""
1707
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1709
def _keys_to_version_ids(self, keys):
1710
return tuple(key[0] for key in keys)
1712
def get_position(self, version_id):
1713
"""Return details needed to access the version.
1715
:return: a tuple (index, data position, size) to hand to the access
1716
logic to get the record.
1718
node = self._get_node(version_id)
1719
bits = node[2][1:].split(' ')
1720
return node[0], int(bits[0]), int(bits[1])
1722
def get_method(self, version_id):
1723
"""Return compression method of specified version."""
1724
if not self._deltas:
1726
return self._parent_compression(self._get_node(version_id)[3][1])
1728
def _parent_compression(self, reference_list):
1729
# use the second reference list to decide if this is delta'd or not.
1730
if len(reference_list):
1735
def _get_node(self, version_id):
1737
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1739
raise RevisionNotPresent(version_id, self)
1741
def get_options(self, version_id):
1742
"""Return a string represention options.
1746
node = self._get_node(version_id)
1747
if not self._deltas:
1748
options = ['fulltext']
1750
options = [self._parent_compression(node[3][1])]
1751
if node[2][0] == 'N':
1752
options.append('no-eol')
1755
def get_parents(self, version_id):
1756
"""Return parents of specified version ignoring ghosts."""
1757
parents = list(self.iter_parents([version_id]))
1760
raise errors.RevisionNotPresent(version_id, self)
1761
return parents[0][1]
1763
def get_parents_with_ghosts(self, version_id):
1764
"""Return parents of specified version with ghosts."""
1765
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1766
check_present=True))
1767
if not self._parents:
1769
return self._keys_to_version_ids(nodes[0][3][0])
1771
def check_versions_present(self, version_ids):
1772
"""Check that all specified versions are present."""
1773
keys = self._version_ids_to_keys(version_ids)
1774
present = self._present_keys(keys)
1775
missing = keys.difference(present)
1777
raise RevisionNotPresent(missing.pop(), self)
1779
def add_version(self, version_id, options, access_memo, parents):
1780
"""Add a version record to the index."""
1781
return self.add_versions(((version_id, options, access_memo, parents),))
1783
def add_versions(self, versions, random_id=False):
1784
"""Add multiple versions to the index.
1786
This function does not insert data into the Immutable GraphIndex
1787
backing the KnitGraphIndex, instead it prepares data for insertion by
1788
the caller and checks that it is safe to insert then calls
1789
self._add_callback with the prepared GraphIndex nodes.
1791
:param versions: a list of tuples:
1792
(version_id, options, pos, size, parents).
1793
:param random_id: If True the ids being added were randomly generated
1794
and no check for existence will be performed.
1796
if not self._add_callback:
1797
raise errors.ReadOnlyError(self)
1798
# we hope there are no repositories with inconsistent parentage
1803
for (version_id, options, access_memo, parents) in versions:
1804
index, pos, size = access_memo
1805
key = (version_id, )
1806
parents = tuple((parent, ) for parent in parents)
1807
if 'no-eol' in options:
1811
value += "%d %d" % (pos, size)
1812
if not self._deltas:
1813
if 'line-delta' in options:
1814
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1817
if 'line-delta' in options:
1818
node_refs = (parents, (parents[0],))
1820
node_refs = (parents, ())
1822
node_refs = (parents, )
1825
raise KnitCorrupt(self, "attempt to add node with parents "
1826
"in parentless index.")
1828
keys[key] = (value, node_refs)
1830
present_nodes = self._get_entries(keys)
1831
for (index, key, value, node_refs) in present_nodes:
1832
if (value, node_refs) != keys[key]:
1833
raise KnitCorrupt(self, "inconsistent details in add_versions"
1834
": %s %s" % ((value, node_refs), keys[key]))
1838
for key, (value, node_refs) in keys.iteritems():
1839
result.append((key, value, node_refs))
1841
for key, (value, node_refs) in keys.iteritems():
1842
result.append((key, value))
1843
self._add_callback(result)
1845
def _version_ids_to_keys(self, version_ids):
1846
return set((version_id, ) for version_id in version_ids)
1849
class _KnitAccess(object):
1850
"""Access to knit records in a .knit file."""
1852
def __init__(self, transport, filename, _file_mode, _dir_mode,
1853
_need_to_create, _create_parent_dir):
1854
"""Create a _KnitAccess for accessing and inserting data.
1856
:param transport: The transport the .knit is located on.
1857
:param filename: The filename of the .knit.
1859
self._transport = transport
1860
self._filename = filename
1861
self._file_mode = _file_mode
1862
self._dir_mode = _dir_mode
1863
self._need_to_create = _need_to_create
1864
self._create_parent_dir = _create_parent_dir
1866
def add_raw_records(self, sizes, raw_data):
1867
"""Add raw knit bytes to a storage area.
1869
The data is spooled to whereever the access method is storing data.
1871
:param sizes: An iterable containing the size of each raw data segment.
1872
:param raw_data: A bytestring containing the data.
1873
:return: A list of memos to retrieve the record later. Each memo is a
1874
tuple - (index, pos, length), where the index field is always None
1875
for the .knit access method.
1877
assert type(raw_data) == str, \
1878
'data must be plain bytes was %s' % type(raw_data)
1879
if not self._need_to_create:
1880
base = self._transport.append_bytes(self._filename, raw_data)
1882
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1883
create_parent_dir=self._create_parent_dir,
1884
mode=self._file_mode,
1885
dir_mode=self._dir_mode)
1886
self._need_to_create = False
1890
result.append((None, base, size))
1895
"""IFF this data access has its own storage area, initialise it.
1899
self._transport.put_bytes_non_atomic(self._filename, '',
1900
mode=self._file_mode)
1902
def open_file(self):
1903
"""IFF this data access can be represented as a single file, open it.
1905
For knits that are not mapped to a single file on disk this will
1908
:return: None or a file handle.
1911
return self._transport.get(self._filename)
1916
def get_raw_records(self, memos_for_retrieval):
1917
"""Get the raw bytes for a records.
1919
:param memos_for_retrieval: An iterable containing the (index, pos,
1920
length) memo for retrieving the bytes. The .knit method ignores
1921
the index as there is always only a single file.
1922
:return: An iterator over the bytes of the records.
1924
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1925
for pos, data in self._transport.readv(self._filename, read_vector):
1929
class _PackAccess(object):
1930
"""Access to knit records via a collection of packs."""
1932
def __init__(self, index_to_packs, writer=None):
1933
"""Create a _PackAccess object.
1935
:param index_to_packs: A dict mapping index objects to the transport
1936
and file names for obtaining data.
1937
:param writer: A tuple (pack.ContainerWriter, write_index) which
1938
contains the pack to write, and the index that reads from it will
1942
self.container_writer = writer[0]
1943
self.write_index = writer[1]
1945
self.container_writer = None
1946
self.write_index = None
1947
self.indices = index_to_packs
1949
def add_raw_records(self, sizes, raw_data):
1950
"""Add raw knit bytes to a storage area.
1952
The data is spooled to the container writer in one bytes-record per
1955
:param sizes: An iterable containing the size of each raw data segment.
1956
:param raw_data: A bytestring containing the data.
1957
:return: A list of memos to retrieve the record later. Each memo is a
1958
tuple - (index, pos, length), where the index field is the
1959
write_index object supplied to the PackAccess object.
1961
assert type(raw_data) == str, \
1962
'data must be plain bytes was %s' % type(raw_data)
1966
p_offset, p_length = self.container_writer.add_bytes_record(
1967
raw_data[offset:offset+size], [])
1969
result.append((self.write_index, p_offset, p_length))
1973
"""Pack based knits do not get individually created."""
1975
def get_raw_records(self, memos_for_retrieval):
1976
"""Get the raw bytes for a records.
1978
:param memos_for_retrieval: An iterable containing the (index, pos,
1979
length) memo for retrieving the bytes. The Pack access method
1980
looks up the pack to use for a given record in its index_to_pack
1982
:return: An iterator over the bytes of the records.
1984
# first pass, group into same-index requests
1986
current_index = None
1987
for (index, offset, length) in memos_for_retrieval:
1988
if current_index == index:
1989
current_list.append((offset, length))
1991
if current_index is not None:
1992
request_lists.append((current_index, current_list))
1993
current_index = index
1994
current_list = [(offset, length)]
1995
# handle the last entry
1996
if current_index is not None:
1997
request_lists.append((current_index, current_list))
1998
for index, offsets in request_lists:
1999
transport, path = self.indices[index]
2000
reader = pack.make_readv_reader(transport, path, offsets)
2001
for names, read_func in reader.iter_records():
2002
yield read_func(None)
2004
def open_file(self):
2005
"""Pack based knits have no single file."""
2008
def set_writer(self, writer, index, (transport, packname)):
2009
"""Set a writer to use for adding data."""
2010
if index is not None:
2011
self.indices[index] = (transport, packname)
2012
self.container_writer = writer
2013
self.write_index = index
2016
class _KnitData(object):
2017
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2019
The KnitData class provides the logic for parsing and using knit records,
2020
making use of an access method for the low level read and write operations.
2023
def __init__(self, access):
2024
"""Create a KnitData object.
2026
:param access: The access method to use. Access methods such as
2027
_KnitAccess manage the insertion of raw records and the subsequent
2028
retrieval of the same.
2030
self._access = access
2031
self._checked = False
2032
# TODO: jam 20060713 conceptually, this could spill to disk
2033
# if the cached size gets larger than a certain amount
2034
# but it complicates the model a bit, so for now just use
2035
# a simple dictionary
2037
self._do_cache = False
2039
def enable_cache(self):
2040
"""Enable caching of reads."""
2041
self._do_cache = True
2043
def clear_cache(self):
2044
"""Clear the record cache."""
2045
self._do_cache = False
2048
def _open_file(self):
2049
return self._access.open_file()
2051
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2052
"""Convert version_id, digest, lines into a raw data block.
2054
:param dense_lines: The bytes of lines but in a denser form. For
2055
instance, if lines is a list of 1000 bytestrings each ending in \n,
2056
dense_lines may be a list with one line in it, containing all the
2057
1000's lines and their \n's. Using dense_lines if it is already
2058
known is a win because the string join to create bytes in this
2059
function spends less time resizing the final string.
2060
:return: (len, a StringIO instance with the raw data ready to read.)
2062
# Note: using a string copy here increases memory pressure with e.g.
2063
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2064
# when doing the initial commit of a mozilla tree. RBC 20070921
2065
bytes = ''.join(chain(
2066
["version %s %d %s\n" % (version_id,
2069
dense_lines or lines,
2070
["end %s\n" % version_id]))
2071
assert bytes.__class__ == str
2072
compressed_bytes = bytes_to_gzip(bytes)
2073
return len(compressed_bytes), compressed_bytes
2075
def add_raw_records(self, sizes, raw_data):
2076
"""Append a prepared record to the data file.
2078
:param sizes: An iterable containing the size of each raw data segment.
2079
:param raw_data: A bytestring containing the data.
2080
:return: a list of index data for the way the data was stored.
2081
See the access method add_raw_records documentation for more
2084
return self._access.add_raw_records(sizes, raw_data)
2086
def _parse_record_header(self, version_id, raw_data):
2087
"""Parse a record header for consistency.
2089
:return: the header and the decompressor stream.
2090
as (stream, header_record)
2092
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2094
rec = self._check_header(version_id, df.readline())
2095
except Exception, e:
2096
raise KnitCorrupt(self._access,
2097
"While reading {%s} got %s(%s)"
2098
% (version_id, e.__class__.__name__, str(e)))
2101
def _check_header(self, version_id, line):
2104
raise KnitCorrupt(self._access,
2105
'unexpected number of elements in record header')
2106
if rec[1] != version_id:
2107
raise KnitCorrupt(self._access,
2108
'unexpected version, wanted %r, got %r'
2109
% (version_id, rec[1]))
2112
def _parse_record(self, version_id, data):
2114
# 4168 calls in 2880 217 internal
2115
# 4168 calls to _parse_record_header in 2121
2116
# 4168 calls to readlines in 330
2117
df = GzipFile(mode='rb', fileobj=StringIO(data))
2120
record_contents = df.readlines()
2121
except Exception, e:
2122
raise KnitCorrupt(self._access,
2123
"While reading {%s} got %s(%s)"
2124
% (version_id, e.__class__.__name__, str(e)))
2125
header = record_contents.pop(0)
2126
rec = self._check_header(version_id, header)
2128
last_line = record_contents.pop()
2129
if len(record_contents) != int(rec[2]):
2130
raise KnitCorrupt(self._access,
2131
'incorrect number of lines %s != %s'
2133
% (len(record_contents), int(rec[2]),
2135
if last_line != 'end %s\n' % rec[1]:
2136
raise KnitCorrupt(self._access,
2137
'unexpected version end line %r, wanted %r'
2138
% (last_line, version_id))
2140
return record_contents, rec[3]
2142
def read_records_iter_raw(self, records):
2143
"""Read text records from data file and yield raw data.
2145
This unpacks enough of the text record to validate the id is
2146
as expected but thats all.
2148
# setup an iterator of the external records:
2149
# uses readv so nice and fast we hope.
2151
# grab the disk data needed.
2153
# Don't check _cache if it is empty
2154
needed_offsets = [index_memo for version_id, index_memo
2156
if version_id not in self._cache]
2158
needed_offsets = [index_memo for version_id, index_memo
2161
raw_records = self._access.get_raw_records(needed_offsets)
2163
for version_id, index_memo in records:
2164
if version_id in self._cache:
2165
# This data has already been validated
2166
data = self._cache[version_id]
2168
data = raw_records.next()
2170
self._cache[version_id] = data
2172
# validate the header
2173
df, rec = self._parse_record_header(version_id, data)
2175
yield version_id, data
2177
def read_records_iter(self, records):
2178
"""Read text records from data file and yield result.
2180
The result will be returned in whatever is the fastest to read.
2181
Not by the order requested. Also, multiple requests for the same
2182
record will only yield 1 response.
2183
:param records: A list of (version_id, pos, len) entries
2184
:return: Yields (version_id, contents, digest) in the order
2185
read, not the order requested
2191
# Skip records we have alread seen
2192
yielded_records = set()
2193
needed_records = set()
2194
for record in records:
2195
if record[0] in self._cache:
2196
if record[0] in yielded_records:
2198
yielded_records.add(record[0])
2199
data = self._cache[record[0]]
2200
content, digest = self._parse_record(record[0], data)
2201
yield (record[0], content, digest)
2203
needed_records.add(record)
2204
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2206
needed_records = sorted(set(records), key=operator.itemgetter(1))
2208
if not needed_records:
2211
# The transport optimizes the fetching as well
2212
# (ie, reads continuous ranges.)
2213
raw_data = self._access.get_raw_records(
2214
[index_memo for version_id, index_memo in needed_records])
2216
for (version_id, index_memo), data in \
2217
izip(iter(needed_records), raw_data):
2218
content, digest = self._parse_record(version_id, data)
2220
self._cache[version_id] = data
2221
yield version_id, content, digest
2223
def read_records(self, records):
2224
"""Read records into a dictionary."""
2226
for record_id, content, digest in \
2227
self.read_records_iter(records):
2228
components[record_id] = (content, digest)
2232
class InterKnit(InterVersionedFile):
2233
"""Optimised code paths for knit to knit operations."""
2235
_matching_file_from_factory = KnitVersionedFile
2236
_matching_file_to_factory = KnitVersionedFile
2239
def is_compatible(source, target):
2240
"""Be compatible with knits. """
2242
return (isinstance(source, KnitVersionedFile) and
2243
isinstance(target, KnitVersionedFile))
2244
except AttributeError:
2247
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2248
"""Copy texts to the target by extracting and adding them one by one.
2250
see join() for the parameter definitions.
2252
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2253
graph = self.source.get_graph(version_ids)
2254
order = topo_sort(graph.items())
2256
def size_of_content(content):
2257
return sum(len(line) for line in content.text())
2258
# Cache at most 10MB of parent texts
2259
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2260
compute_size=size_of_content)
2261
# TODO: jam 20071116 It would be nice to have a streaming interface to
2262
# get multiple texts from a source. The source could be smarter
2263
# about how it handled intermediate stages.
2264
# get_line_list() or make_mpdiffs() seem like a possibility, but
2265
# at the moment they extract all full texts into memory, which
2266
# causes us to store more than our 3x fulltext goal.
2267
# Repository.iter_files_bytes() may be another possibility
2268
to_process = [version for version in order
2269
if version not in self.target]
2270
total = len(to_process)
2271
pb = ui.ui_factory.nested_progress_bar()
2273
for index, version in enumerate(to_process):
2274
pb.update('Converting versioned data', index, total)
2275
sha1, num_bytes, parent_text = self.target.add_lines(version,
2276
self.source.get_parents(version),
2277
self.source.get_lines(version),
2278
parent_texts=parent_cache)
2279
parent_cache[version] = parent_text
2284
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2285
"""See InterVersionedFile.join."""
2286
assert isinstance(self.source, KnitVersionedFile)
2287
assert isinstance(self.target, KnitVersionedFile)
2289
# If the source and target are mismatched w.r.t. annotations vs
2290
# plain, the data needs to be converted accordingly
2291
if self.source.factory.annotated == self.target.factory.annotated:
2293
elif self.source.factory.annotated:
2294
converter = self._anno_to_plain_converter
2296
# We're converting from a plain to an annotated knit. Copy them
2297
# across by full texts.
2298
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2300
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2304
pb = ui.ui_factory.nested_progress_bar()
2306
version_ids = list(version_ids)
2307
if None in version_ids:
2308
version_ids.remove(None)
2310
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2311
this_versions = set(self.target._index.get_versions())
2312
# XXX: For efficiency we should not look at the whole index,
2313
# we only need to consider the referenced revisions - they
2314
# must all be present, or the method must be full-text.
2315
# TODO, RBC 20070919
2316
needed_versions = self.source_ancestry - this_versions
2318
if not needed_versions:
2320
full_list = topo_sort(self.source.get_graph())
2322
version_list = [i for i in full_list if (not self.target.has_version(i)
2323
and i in needed_versions)]
2327
copy_queue_records = []
2329
for version_id in version_list:
2330
options = self.source._index.get_options(version_id)
2331
parents = self.source._index.get_parents_with_ghosts(version_id)
2332
# check that its will be a consistent copy:
2333
for parent in parents:
2334
# if source has the parent, we must :
2335
# * already have it or
2336
# * have it scheduled already
2337
# otherwise we don't care
2338
assert (self.target.has_version(parent) or
2339
parent in copy_set or
2340
not self.source.has_version(parent))
2341
index_memo = self.source._index.get_position(version_id)
2342
copy_queue_records.append((version_id, index_memo))
2343
copy_queue.append((version_id, options, parents))
2344
copy_set.add(version_id)
2346
# data suck the join:
2348
total = len(version_list)
2351
for (version_id, raw_data), \
2352
(version_id2, options, parents) in \
2353
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2355
assert version_id == version_id2, 'logic error, inconsistent results'
2357
pb.update("Joining knit", count, total)
2359
size, raw_data = converter(raw_data, version_id, options,
2362
size = len(raw_data)
2363
raw_records.append((version_id, options, parents, size))
2364
raw_datum.append(raw_data)
2365
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2370
def _anno_to_plain_converter(self, raw_data, version_id, options,
2372
"""Convert annotated content to plain content."""
2373
data, digest = self.source._data._parse_record(version_id, raw_data)
2374
if 'fulltext' in options:
2375
content = self.source.factory.parse_fulltext(data, version_id)
2376
lines = self.target.factory.lower_fulltext(content)
2378
delta = self.source.factory.parse_line_delta(data, version_id,
2380
lines = self.target.factory.lower_line_delta(delta)
2381
return self.target._data._record_to_data(version_id, digest, lines)
2384
InterVersionedFile.register_optimiser(InterKnit)
2387
class WeaveToKnit(InterVersionedFile):
2388
"""Optimised code paths for weave to knit operations."""
2390
_matching_file_from_factory = bzrlib.weave.WeaveFile
2391
_matching_file_to_factory = KnitVersionedFile
2394
def is_compatible(source, target):
2395
"""Be compatible with weaves to knits."""
2397
return (isinstance(source, bzrlib.weave.Weave) and
2398
isinstance(target, KnitVersionedFile))
2399
except AttributeError:
2402
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2403
"""See InterVersionedFile.join."""
2404
assert isinstance(self.source, bzrlib.weave.Weave)
2405
assert isinstance(self.target, KnitVersionedFile)
2407
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2412
pb = ui.ui_factory.nested_progress_bar()
2414
version_ids = list(version_ids)
2416
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2417
this_versions = set(self.target._index.get_versions())
2418
needed_versions = self.source_ancestry - this_versions
2420
if not needed_versions:
2422
full_list = topo_sort(self.source.get_graph())
2424
version_list = [i for i in full_list if (not self.target.has_version(i)
2425
and i in needed_versions)]
2429
total = len(version_list)
2430
for version_id in version_list:
2431
pb.update("Converting to knit", count, total)
2432
parents = self.source.get_parents(version_id)
2433
# check that its will be a consistent copy:
2434
for parent in parents:
2435
# if source has the parent, we must already have it
2436
assert (self.target.has_version(parent))
2437
self.target.add_lines(
2438
version_id, parents, self.source.get_lines(version_id))
2445
InterVersionedFile.register_optimiser(WeaveToKnit)
2448
# Deprecated, use PatienceSequenceMatcher instead
2449
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2452
def annotate_knit(knit, revision_id):
2453
"""Annotate a knit with no cached annotations.
2455
This implementation is for knits with no cached annotations.
2456
It will work for knits with cached annotations, but this is not
2459
ancestry = knit.get_ancestry(revision_id)
2460
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2462
for candidate in ancestry:
2463
if candidate in annotations:
2465
parents = knit.get_parents(candidate)
2466
if len(parents) == 0:
2468
elif knit._index.get_method(candidate) != 'line-delta':
2471
parent, sha1, noeol, delta = knit.get_delta(candidate)
2472
blocks = KnitContent.get_line_delta_blocks(delta,
2473
fulltext[parents[0]], fulltext[candidate])
2474
annotations[candidate] = list(annotate.reannotate([annotations[p]
2475
for p in parents], fulltext[candidate], candidate, blocks))
2476
return iter(annotations[revision_id])
2480
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2482
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
