1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
"""Knit versionedfile implementation.
19
A knit is a versioned file implementation that supports efficient append only
23
lifeless: the data file is made up of "delta records". each delta record has a delta header
24
that contains; (1) a version id, (2) the size of the delta (in lines), and (3) the digest of
25
the -expanded data- (ie, the delta applied to the parent). the delta also ends with a
26
end-marker; simply "end VERSION"
28
delta can be line or full contents.a
29
... the 8's there are the index number of the annotation.
30
version robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad 7 c7d23b2a5bd6ca00e8e266cec0ec228158ee9f9e
34
8 e.set('executable', 'yes')
36
8 if elt.get('executable') == 'yes':
37
8 ie.executable = True
38
end robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad
42
09:33 < jrydberg> lifeless: each index is made up of a tuple of; version id, options, position, size, parents
43
09:33 < jrydberg> lifeless: the parents are currently dictionary compressed
44
09:33 < jrydberg> lifeless: (meaning it currently does not support ghosts)
45
09:33 < lifeless> right
46
09:33 < jrydberg> lifeless: the position and size is the range in the data file
49
so the index sequence is the dictionary compressed sequence number used
50
in the deltas to provide line annotation
55
# 10:16 < lifeless> make partial index writes safe
56
# 10:16 < lifeless> implement 'knit.check()' like weave.check()
57
# 10:17 < lifeless> record known ghosts so we can detect when they are filled in rather than the current 'reweave
59
# move sha1 out of the content so that join is faster at verifying parents
60
# record content length ?
64
from cStringIO import StringIO
65
from itertools import izip, chain
70
from zlib import Z_DEFAULT_COMPRESSION
73
from bzrlib.lazy_import import lazy_import
74
lazy_import(globals(), """
93
from bzrlib.errors import (
99
KnitDataStreamIncompatible,
102
RevisionAlreadyPresent,
104
from bzrlib.tuned_gzip import GzipFile, bytes_to_gzip
105
from bzrlib.osutils import (
111
from bzrlib.symbol_versioning import DEPRECATED_PARAMETER, deprecated_passed
112
from bzrlib.tsort import topo_sort
115
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
118
# TODO: Split out code specific to this format into an associated object.
120
# TODO: Can we put in some kind of value to check that the index and data
121
# files belong together?
123
# TODO: accommodate binaries, perhaps by storing a byte count
125
# TODO: function to check whole file
127
# TODO: atomically append data, then measure backwards from the cursor
128
# position after writing to work out where it was located. we may need to
129
# bypass python file buffering.
131
DATA_SUFFIX = '.knit'
132
INDEX_SUFFIX = '.kndx'
135
class KnitContent(object):
136
"""Content of a knit version to which deltas can be applied."""
139
"""Return a list of (origin, text) tuples."""
140
return list(self.annotate_iter())
142
def apply_delta(self, delta, new_version_id):
143
"""Apply delta to this object to become new_version_id."""
144
raise NotImplementedError(self.apply_delta)
146
def line_delta_iter(self, new_lines):
147
"""Generate line-based delta from this content to new_lines."""
148
new_texts = new_lines.text()
149
old_texts = self.text()
150
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
151
for tag, i1, i2, j1, j2 in s.get_opcodes():
154
# ofrom, oto, length, data
155
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
157
def line_delta(self, new_lines):
158
return list(self.line_delta_iter(new_lines))
161
def get_line_delta_blocks(knit_delta, source, target):
162
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
163
target_len = len(target)
166
for s_begin, s_end, t_len, new_text in knit_delta:
167
true_n = s_begin - s_pos
170
# knit deltas do not provide reliable info about whether the
171
# last line of a file matches, due to eol handling.
172
if source[s_pos + n -1] != target[t_pos + n -1]:
175
yield s_pos, t_pos, n
176
t_pos += t_len + true_n
178
n = target_len - t_pos
180
if source[s_pos + n -1] != target[t_pos + n -1]:
183
yield s_pos, t_pos, n
184
yield s_pos + (target_len - t_pos), target_len, 0
187
class AnnotatedKnitContent(KnitContent):
188
"""Annotated content."""
190
def __init__(self, lines):
193
def annotate_iter(self):
194
"""Yield tuples of (origin, text) for each content line."""
195
return iter(self._lines)
197
def apply_delta(self, delta, new_version_id):
198
"""Apply delta to this object to become new_version_id."""
201
for start, end, count, delta_lines in delta:
202
lines[offset+start:offset+end] = delta_lines
203
offset = offset + (start - end) + count
205
def strip_last_line_newline(self):
206
line = self._lines[-1][1].rstrip('\n')
207
self._lines[-1] = (self._lines[-1][0], line)
211
return [text for origin, text in self._lines]
212
except ValueError, e:
213
# most commonly (only?) caused by the internal form of the knit
214
# missing annotation information because of a bug - see thread
216
raise KnitCorrupt(self,
217
"line in annotated knit missing annotation information: %s"
221
return AnnotatedKnitContent(self._lines[:])
224
class PlainKnitContent(KnitContent):
225
"""Unannotated content.
227
When annotate[_iter] is called on this content, the same version is reported
228
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
232
def __init__(self, lines, version_id):
234
self._version_id = version_id
236
def annotate_iter(self):
237
"""Yield tuples of (origin, text) for each content line."""
238
for line in self._lines:
239
yield self._version_id, line
241
def apply_delta(self, delta, new_version_id):
242
"""Apply delta to this object to become new_version_id."""
245
for start, end, count, delta_lines in delta:
246
lines[offset+start:offset+end] = delta_lines
247
offset = offset + (start - end) + count
248
self._version_id = new_version_id
251
return PlainKnitContent(self._lines[:], self._version_id)
253
def strip_last_line_newline(self):
254
self._lines[-1] = self._lines[-1].rstrip('\n')
260
class KnitAnnotateFactory(object):
261
"""Factory for creating annotated Content objects."""
265
def make(self, lines, version_id):
266
num_lines = len(lines)
267
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
269
def parse_fulltext(self, content, version_id):
270
"""Convert fulltext to internal representation
272
fulltext content is of the format
273
revid(utf8) plaintext\n
274
internal representation is of the format:
277
# TODO: jam 20070209 The tests expect this to be returned as tuples,
278
# but the code itself doesn't really depend on that.
279
# Figure out a way to not require the overhead of turning the
280
# list back into tuples.
281
lines = [tuple(line.split(' ', 1)) for line in content]
282
return AnnotatedKnitContent(lines)
284
def parse_line_delta_iter(self, lines):
285
return iter(self.parse_line_delta(lines))
287
def parse_line_delta(self, lines, version_id, plain=False):
288
"""Convert a line based delta into internal representation.
290
line delta is in the form of:
291
intstart intend intcount
293
revid(utf8) newline\n
294
internal representation is
295
(start, end, count, [1..count tuples (revid, newline)])
297
:param plain: If True, the lines are returned as a plain
298
list without annotations, not as a list of (origin, content) tuples, i.e.
299
(start, end, count, [1..count newline])
306
def cache_and_return(line):
307
origin, text = line.split(' ', 1)
308
return cache.setdefault(origin, origin), text
310
# walk through the lines parsing.
311
# Note that the plain test is explicitly pulled out of the
312
# loop to minimise any performance impact
315
start, end, count = [int(n) for n in header.split(',')]
316
contents = [next().split(' ', 1)[1] for i in xrange(count)]
317
result.append((start, end, count, contents))
320
start, end, count = [int(n) for n in header.split(',')]
321
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
322
result.append((start, end, count, contents))
325
def get_fulltext_content(self, lines):
326
"""Extract just the content lines from a fulltext."""
327
return (line.split(' ', 1)[1] for line in lines)
329
def get_linedelta_content(self, lines):
330
"""Extract just the content from a line delta.
332
This doesn't return all of the extra information stored in a delta.
333
Only the actual content lines.
338
header = header.split(',')
339
count = int(header[2])
340
for i in xrange(count):
341
origin, text = next().split(' ', 1)
344
def lower_fulltext(self, content):
345
"""convert a fulltext content record into a serializable form.
347
see parse_fulltext which this inverts.
349
# TODO: jam 20070209 We only do the caching thing to make sure that
350
# the origin is a valid utf-8 line, eventually we could remove it
351
return ['%s %s' % (o, t) for o, t in content._lines]
353
def lower_line_delta(self, delta):
354
"""convert a delta into a serializable form.
356
See parse_line_delta which this inverts.
358
# TODO: jam 20070209 We only do the caching thing to make sure that
359
# the origin is a valid utf-8 line, eventually we could remove it
361
for start, end, c, lines in delta:
362
out.append('%d,%d,%d\n' % (start, end, c))
363
out.extend(origin + ' ' + text
364
for origin, text in lines)
367
def annotate_iter(self, knit, version_id):
368
content = knit._get_content(version_id)
369
return content.annotate_iter()
372
class KnitPlainFactory(object):
373
"""Factory for creating plain Content objects."""
377
def make(self, lines, version_id):
378
return PlainKnitContent(lines, version_id)
380
def parse_fulltext(self, content, version_id):
381
"""This parses an unannotated fulltext.
383
Note that this is not a noop - the internal representation
384
has (versionid, line) - its just a constant versionid.
386
return self.make(content, version_id)
388
def parse_line_delta_iter(self, lines, version_id):
390
num_lines = len(lines)
391
while cur < num_lines:
394
start, end, c = [int(n) for n in header.split(',')]
395
yield start, end, c, lines[cur:cur+c]
398
def parse_line_delta(self, lines, version_id):
399
return list(self.parse_line_delta_iter(lines, version_id))
401
def get_fulltext_content(self, lines):
402
"""Extract just the content lines from a fulltext."""
405
def get_linedelta_content(self, lines):
406
"""Extract just the content from a line delta.
408
This doesn't return all of the extra information stored in a delta.
409
Only the actual content lines.
414
header = header.split(',')
415
count = int(header[2])
416
for i in xrange(count):
419
def lower_fulltext(self, content):
420
return content.text()
422
def lower_line_delta(self, delta):
424
for start, end, c, lines in delta:
425
out.append('%d,%d,%d\n' % (start, end, c))
429
def annotate_iter(self, knit, version_id):
430
return annotate_knit(knit, version_id)
433
def make_empty_knit(transport, relpath):
434
"""Construct a empty knit at the specified location."""
435
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
438
class KnitVersionedFile(VersionedFile):
439
"""Weave-like structure with faster random access.
441
A knit stores a number of texts and a summary of the relationships
442
between them. Texts are identified by a string version-id. Texts
443
are normally stored and retrieved as a series of lines, but can
444
also be passed as single strings.
446
Lines are stored with the trailing newline (if any) included, to
447
avoid special cases for files with no final newline. Lines are
448
composed of 8-bit characters, not unicode. The combination of
449
these approaches should mean any 'binary' file can be safely
450
stored and retrieved.
453
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
454
factory=None, delta=True, create=False, create_parent_dir=False,
455
delay_create=False, dir_mode=None, index=None, access_method=None):
456
"""Construct a knit at location specified by relpath.
458
:param create: If not True, only open an existing knit.
459
:param create_parent_dir: If True, create the parent directory if
460
creating the file fails. (This is used for stores with
461
hash-prefixes that may not exist yet)
462
:param delay_create: The calling code is aware that the knit won't
463
actually be created until the first data is stored.
464
:param index: An index to use for the knit.
466
if access_mode is None:
468
super(KnitVersionedFile, self).__init__(access_mode)
469
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
470
self.transport = transport
471
self.filename = relpath
472
self.factory = factory or KnitAnnotateFactory()
473
self.writable = (access_mode == 'w')
476
self._max_delta_chain = 200
479
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
480
access_mode, create=create, file_mode=file_mode,
481
create_parent_dir=create_parent_dir, delay_create=delay_create,
485
if access_method is None:
486
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
487
((create and not len(self)) and delay_create), create_parent_dir)
489
_access = access_method
490
if create and not len(self) and not delay_create:
492
self._data = _KnitData(_access)
495
return '%s(%s)' % (self.__class__.__name__,
496
self.transport.abspath(self.filename))
498
def _check_should_delta(self, first_parents):
499
"""Iterate back through the parent listing, looking for a fulltext.
501
This is used when we want to decide whether to add a delta or a new
502
fulltext. It searches for _max_delta_chain parents. When it finds a
503
fulltext parent, it sees if the total size of the deltas leading up to
504
it is large enough to indicate that we want a new full text anyway.
506
Return True if we should create a new delta, False if we should use a
511
delta_parents = first_parents
512
for count in xrange(self._max_delta_chain):
513
parent = delta_parents[0]
514
method = self._index.get_method(parent)
515
index, pos, size = self._index.get_position(parent)
516
if method == 'fulltext':
520
delta_parents = self._index.get_parents(parent)
522
# We couldn't find a fulltext, so we must create a new one
525
return fulltext_size > delta_size
527
def _add_raw_records(self, records, data):
528
"""Add all the records 'records' with data pre-joined in 'data'.
530
:param records: A list of tuples(version_id, options, parents, size).
531
:param data: The data for the records. When it is written, the records
532
are adjusted to have pos pointing into data by the sum of
533
the preceding records sizes.
536
raw_record_sizes = [record[3] for record in records]
537
positions = self._data.add_raw_records(raw_record_sizes, data)
540
for (version_id, options, parents, size), access_memo in zip(
542
index_entries.append((version_id, options, access_memo, parents))
543
if self._data._do_cache:
544
self._data._cache[version_id] = data[offset:offset+size]
546
self._index.add_versions(index_entries)
548
def enable_cache(self):
549
"""Start caching data for this knit"""
550
self._data.enable_cache()
552
def clear_cache(self):
553
"""Clear the data cache only."""
554
self._data.clear_cache()
556
def copy_to(self, name, transport):
557
"""See VersionedFile.copy_to()."""
558
# copy the current index to a temp index to avoid racing with local
560
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
561
self.transport.get(self._index._filename))
563
f = self._data._open_file()
565
transport.put_file(name + DATA_SUFFIX, f)
568
# move the copied index into place
569
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
571
def create_empty(self, name, transport, mode=None):
572
return KnitVersionedFile(name, transport, factory=self.factory,
573
delta=self.delta, create=True)
575
def get_data_stream(self, required_versions):
576
"""Get a data stream for the specified versions.
578
Versions may be returned in any order, not necessarily the order
579
specified. They are returned in a partial order by compression
580
parent, so that the deltas can be applied as the data stream is
581
inserted; however note that compression parents will not be sent
582
unless they were specifically requested, as the client may already
585
:param required_versions: The exact set of versions to be extracted.
586
Unlike some other knit methods, this is not used to generate a
587
transitive closure, rather it is used precisely as given.
589
:returns: format_signature, list of (version, options, length, parents),
592
required_version_set = frozenset(required_versions)
594
# list of revisions that can just be sent without waiting for their
597
# map from revision to the children based on it
599
# first, read all relevant index data, enough to sort into the right
601
for version_id in required_versions:
602
options = self._index.get_options(version_id)
603
parents = self._index.get_parents_with_ghosts(version_id)
604
index_memo = self._index.get_position(version_id)
605
version_index[version_id] = (index_memo, options, parents)
606
if ('line-delta' in options
607
and parents[0] in required_version_set):
608
# must wait until the parent has been sent
609
deferred.setdefault(parents[0], []). \
612
# either a fulltext, or a delta whose parent the client did
613
# not ask for and presumably already has
614
ready_to_send.append(version_id)
615
# build a list of results to return, plus instructions for data to
617
copy_queue_records = []
618
temp_version_list = []
620
# XXX: pushing and popping lists may be a bit inefficient
621
version_id = ready_to_send.pop(0)
622
(index_memo, options, parents) = version_index[version_id]
623
copy_queue_records.append((version_id, index_memo))
624
none, data_pos, data_size = index_memo
625
temp_version_list.append((version_id, options, data_size,
627
if version_id in deferred:
628
# now we can send all the children of this revision - we could
629
# put them in anywhere, but we hope that sending them soon
630
# after the fulltext will give good locality in the receiver
631
ready_to_send[:0] = deferred.pop(version_id)
632
assert len(deferred) == 0, \
633
"Still have compressed child versions waiting to be sent"
634
# XXX: The stream format is such that we cannot stream it - we have to
635
# know the length of all the data a-priori.
637
result_version_list = []
638
for (version_id, raw_data), \
639
(version_id2, options, _, parents) in \
640
izip(self._data.read_records_iter_raw(copy_queue_records),
642
assert version_id == version_id2, \
643
'logic error, inconsistent results'
644
raw_datum.append(raw_data)
645
result_version_list.append(
646
(version_id, options, len(raw_data), parents))
647
# provide a callback to get data incrementally.
648
pseudo_file = StringIO(''.join(raw_datum))
651
return pseudo_file.read()
653
return pseudo_file.read(length)
654
return (self.get_format_signature(), result_version_list, read)
656
def _extract_blocks(self, version_id, source, target):
657
if self._index.get_method(version_id) != 'line-delta':
659
parent, sha1, noeol, delta = self.get_delta(version_id)
660
return KnitContent.get_line_delta_blocks(delta, source, target)
662
def get_delta(self, version_id):
663
"""Get a delta for constructing version from some other version."""
664
self.check_not_reserved_id(version_id)
665
parents = self.get_parents(version_id)
670
index_memo = self._index.get_position(version_id)
671
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
672
noeol = 'no-eol' in self._index.get_options(version_id)
673
if 'fulltext' == self._index.get_method(version_id):
674
new_content = self.factory.parse_fulltext(data, version_id)
675
if parent is not None:
676
reference_content = self._get_content(parent)
677
old_texts = reference_content.text()
680
new_texts = new_content.text()
681
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
683
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
685
delta = self.factory.parse_line_delta(data, version_id)
686
return parent, sha1, noeol, delta
688
def get_format_signature(self):
689
"""See VersionedFile.get_format_signature()."""
690
if self.factory.annotated:
691
annotated_part = "annotated"
693
annotated_part = "plain"
694
return "knit-%s" % (annotated_part,)
696
def get_graph_with_ghosts(self):
697
"""See VersionedFile.get_graph_with_ghosts()."""
698
graph_items = self._index.get_graph()
699
return dict(graph_items)
701
def get_sha1(self, version_id):
702
return self.get_sha1s([version_id])[0]
704
def get_sha1s(self, version_ids):
705
"""See VersionedFile.get_sha1()."""
706
record_map = self._get_record_map(version_ids)
707
# record entry 2 is the 'digest'.
708
return [record_map[v][2] for v in version_ids]
712
"""See VersionedFile.get_suffixes()."""
713
return [DATA_SUFFIX, INDEX_SUFFIX]
715
def has_ghost(self, version_id):
716
"""True if there is a ghost reference in the file to version_id."""
718
if self.has_version(version_id):
720
# optimisable if needed by memoising the _ghosts set.
721
items = self._index.get_graph()
722
for node, parents in items:
723
for parent in parents:
724
if parent not in self._index._cache:
725
if parent == version_id:
729
def insert_data_stream(self, (format, data_list, reader_callable)):
730
"""Insert knit records from a data stream into this knit.
732
If a version in the stream is already present in this knit, it will not
733
be inserted a second time. It will be checked for consistency with the
734
stored version however, and may cause a KnitCorrupt error to be raised
735
if the data in the stream disagrees with the already stored data.
737
:seealso: get_data_stream
739
if format != self.get_format_signature():
740
trace.mutter('incompatible format signature inserting to %r', self)
741
raise KnitDataStreamIncompatible(
742
format, self.get_format_signature())
744
for version_id, options, length, parents in data_list:
745
if self.has_version(version_id):
746
# First check: the list of parents.
747
my_parents = self.get_parents_with_ghosts(version_id)
748
if my_parents != parents:
749
# XXX: KnitCorrupt is not quite the right exception here.
752
'parents list %r from data stream does not match '
753
'already recorded parents %r for %s'
754
% (parents, my_parents, version_id))
756
# Also check the SHA-1 of the fulltext this content will
758
raw_data = reader_callable(length)
759
my_fulltext_sha1 = self.get_sha1(version_id)
760
df, rec = self._data._parse_record_header(version_id, raw_data)
761
stream_fulltext_sha1 = rec[3]
762
if my_fulltext_sha1 != stream_fulltext_sha1:
763
# Actually, we don't know if it's this knit that's corrupt,
764
# or the data stream we're trying to insert.
766
self.filename, 'sha-1 does not match %s' % version_id)
768
if 'line-delta' in options:
769
# Make sure that this knit record is actually useful: a
770
# line-delta is no use unless we have its parent.
771
# Fetching from a broken repository with this problem
772
# shouldn't break the target repository.
774
# See https://bugs.launchpad.net/bzr/+bug/164443
775
if not self._index.has_version(parents[0]):
778
'line-delta from stream '
781
'missing parent %s\n'
782
'Try running "bzr check" '
783
'on the source repository, and "bzr reconcile" '
785
(version_id, parents[0]))
786
self._add_raw_records(
787
[(version_id, options, parents, length)],
788
reader_callable(length))
791
"""See VersionedFile.versions."""
792
if 'evil' in debug.debug_flags:
793
trace.mutter_callsite(2, "versions scales with size of history")
794
return self._index.get_versions()
796
def has_version(self, version_id):
797
"""See VersionedFile.has_version."""
798
if 'evil' in debug.debug_flags:
799
trace.mutter_callsite(2, "has_version is a LBYL scenario")
800
return self._index.has_version(version_id)
802
__contains__ = has_version
804
def _merge_annotations(self, content, parents, parent_texts={},
805
delta=None, annotated=None,
806
left_matching_blocks=None):
807
"""Merge annotations for content. This is done by comparing
808
the annotations based on changed to the text.
810
if left_matching_blocks is not None:
811
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
815
for parent_id in parents:
816
merge_content = self._get_content(parent_id, parent_texts)
817
if (parent_id == parents[0] and delta_seq is not None):
820
seq = patiencediff.PatienceSequenceMatcher(
821
None, merge_content.text(), content.text())
822
for i, j, n in seq.get_matching_blocks():
825
# this appears to copy (origin, text) pairs across to the
826
# new content for any line that matches the last-checked
828
content._lines[j:j+n] = merge_content._lines[i:i+n]
830
if delta_seq is None:
831
reference_content = self._get_content(parents[0], parent_texts)
832
new_texts = content.text()
833
old_texts = reference_content.text()
834
delta_seq = patiencediff.PatienceSequenceMatcher(
835
None, old_texts, new_texts)
836
return self._make_line_delta(delta_seq, content)
838
def _make_line_delta(self, delta_seq, new_content):
839
"""Generate a line delta from delta_seq and new_content."""
841
for op in delta_seq.get_opcodes():
844
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
847
def _get_components_positions(self, version_ids):
848
"""Produce a map of position data for the components of versions.
850
This data is intended to be used for retrieving the knit records.
852
A dict of version_id to (method, data_pos, data_size, next) is
854
method is the way referenced data should be applied.
855
data_pos is the position of the data in the knit.
856
data_size is the size of the data in the knit.
857
next is the build-parent of the version, or None for fulltexts.
860
for version_id in version_ids:
863
while cursor is not None and cursor not in component_data:
864
method = self._index.get_method(cursor)
865
if method == 'fulltext':
868
next = self.get_parents_with_ghosts(cursor)[0]
869
index_memo = self._index.get_position(cursor)
870
component_data[cursor] = (method, index_memo, next)
872
return component_data
874
def _get_content(self, version_id, parent_texts={}):
875
"""Returns a content object that makes up the specified
877
cached_version = parent_texts.get(version_id, None)
878
if cached_version is not None:
879
if not self.has_version(version_id):
880
raise RevisionNotPresent(version_id, self.filename)
881
return cached_version
883
text_map, contents_map = self._get_content_maps([version_id])
884
return contents_map[version_id]
886
def _check_versions_present(self, version_ids):
887
"""Check that all specified versions are present."""
888
self._index.check_versions_present(version_ids)
890
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
891
nostore_sha, random_id, check_content):
892
"""See VersionedFile.add_lines_with_ghosts()."""
893
self._check_add(version_id, lines, random_id, check_content)
894
return self._add(version_id, lines, parents, self.delta,
895
parent_texts, None, nostore_sha, random_id)
897
def _add_lines(self, version_id, parents, lines, parent_texts,
898
left_matching_blocks, nostore_sha, random_id, check_content):
899
"""See VersionedFile.add_lines."""
900
self._check_add(version_id, lines, random_id, check_content)
901
self._check_versions_present(parents)
902
return self._add(version_id, lines[:], parents, self.delta,
903
parent_texts, left_matching_blocks, nostore_sha, random_id)
905
def _check_add(self, version_id, lines, random_id, check_content):
906
"""check that version_id and lines are safe to add."""
907
if contains_whitespace(version_id):
908
raise InvalidRevisionId(version_id, self.filename)
909
self.check_not_reserved_id(version_id)
910
# Technically this could be avoided if we are happy to allow duplicate
911
# id insertion when other things than bzr core insert texts, but it
912
# seems useful for folk using the knit api directly to have some safety
913
# blanket that we can disable.
914
if not random_id and self.has_version(version_id):
915
raise RevisionAlreadyPresent(version_id, self.filename)
917
self._check_lines_not_unicode(lines)
918
self._check_lines_are_lines(lines)
920
def _add(self, version_id, lines, parents, delta, parent_texts,
921
left_matching_blocks, nostore_sha, random_id):
922
"""Add a set of lines on top of version specified by parents.
924
If delta is true, compress the text as a line-delta against
927
Any versions not present will be converted into ghosts.
929
# first thing, if the content is something we don't need to store, find
931
line_bytes = ''.join(lines)
932
digest = sha_string(line_bytes)
933
if nostore_sha == digest:
934
raise errors.ExistingContent
937
if parent_texts is None:
939
for parent in parents:
940
if self.has_version(parent):
941
present_parents.append(parent)
943
# can only compress against the left most present parent.
945
(len(present_parents) == 0 or
946
present_parents[0] != parents[0])):
949
text_length = len(line_bytes)
952
if lines[-1][-1] != '\n':
953
# copy the contents of lines.
955
options.append('no-eol')
956
lines[-1] = lines[-1] + '\n'
960
# To speed the extract of texts the delta chain is limited
961
# to a fixed number of deltas. This should minimize both
962
# I/O and the time spend applying deltas.
963
delta = self._check_should_delta(present_parents)
965
assert isinstance(version_id, str)
966
content = self.factory.make(lines, version_id)
967
if delta or (self.factory.annotated and len(present_parents) > 0):
968
# Merge annotations from parent texts if needed.
969
delta_hunks = self._merge_annotations(content, present_parents,
970
parent_texts, delta, self.factory.annotated,
971
left_matching_blocks)
974
options.append('line-delta')
975
store_lines = self.factory.lower_line_delta(delta_hunks)
976
size, bytes = self._data._record_to_data(version_id, digest,
979
options.append('fulltext')
980
# isinstance is slower and we have no hierarchy.
981
if self.factory.__class__ == KnitPlainFactory:
982
# Use the already joined bytes saving iteration time in
984
size, bytes = self._data._record_to_data(version_id, digest,
987
# get mixed annotation + content and feed it into the
989
store_lines = self.factory.lower_fulltext(content)
990
size, bytes = self._data._record_to_data(version_id, digest,
993
access_memo = self._data.add_raw_records([size], bytes)[0]
994
self._index.add_versions(
995
((version_id, options, access_memo, parents),),
997
return digest, text_length, content
999
def check(self, progress_bar=None):
1000
"""See VersionedFile.check()."""
1002
def _clone_text(self, new_version_id, old_version_id, parents):
1003
"""See VersionedFile.clone_text()."""
1004
# FIXME RBC 20060228 make fast by only inserting an index with null
1006
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1008
def get_lines(self, version_id):
1009
"""See VersionedFile.get_lines()."""
1010
return self.get_line_list([version_id])[0]
1012
def _get_record_map(self, version_ids):
1013
"""Produce a dictionary of knit records.
1015
The keys are version_ids, the values are tuples of (method, content,
1017
method is the way the content should be applied.
1018
content is a KnitContent object.
1019
digest is the SHA1 digest of this version id after all steps are done
1020
next is the build-parent of the version, i.e. the leftmost ancestor.
1021
If the method is fulltext, next will be None.
1023
position_map = self._get_components_positions(version_ids)
1024
# c = component_id, m = method, i_m = index_memo, n = next
1025
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1027
for component_id, content, digest in \
1028
self._data.read_records_iter(records):
1029
method, index_memo, next = position_map[component_id]
1030
record_map[component_id] = method, content, digest, next
1034
def get_text(self, version_id):
1035
"""See VersionedFile.get_text"""
1036
return self.get_texts([version_id])[0]
1038
def get_texts(self, version_ids):
1039
return [''.join(l) for l in self.get_line_list(version_ids)]
1041
def get_line_list(self, version_ids):
1042
"""Return the texts of listed versions as a list of strings."""
1043
for version_id in version_ids:
1044
self.check_not_reserved_id(version_id)
1045
text_map, content_map = self._get_content_maps(version_ids)
1046
return [text_map[v] for v in version_ids]
1048
_get_lf_split_line_list = get_line_list
1050
def _get_content_maps(self, version_ids):
1051
"""Produce maps of text and KnitContents
1053
:return: (text_map, content_map) where text_map contains the texts for
1054
the requested versions and content_map contains the KnitContents.
1055
Both dicts take version_ids as their keys.
1057
# FUTURE: This function could be improved for the 'extract many' case
1058
# by tracking each component and only doing the copy when the number of
1059
# children than need to apply delta's to it is > 1 or it is part of the
1061
version_ids = list(version_ids)
1062
multiple_versions = len(version_ids) != 1
1063
record_map = self._get_record_map(version_ids)
1068
for version_id in version_ids:
1071
while cursor is not None:
1072
method, data, digest, next = record_map[cursor]
1073
components.append((cursor, method, data, digest))
1074
if cursor in content_map:
1079
for component_id, method, data, digest in reversed(components):
1080
if component_id in content_map:
1081
content = content_map[component_id]
1083
if method == 'fulltext':
1084
assert content is None
1085
content = self.factory.parse_fulltext(data, version_id)
1086
elif method == 'line-delta':
1087
delta = self.factory.parse_line_delta(data, version_id)
1088
if multiple_versions:
1089
# only doing this when we want multiple versions
1090
# output avoids list copies - which reference and
1091
# dereference many strings.
1092
content = content.copy()
1093
content.apply_delta(delta, version_id)
1094
if multiple_versions:
1095
content_map[component_id] = content
1097
if 'no-eol' in self._index.get_options(version_id):
1098
if multiple_versions:
1099
content = content.copy()
1100
content.strip_last_line_newline()
1101
final_content[version_id] = content
1103
# digest here is the digest from the last applied component.
1104
text = content.text()
1105
actual_sha = sha_strings(text)
1106
if actual_sha != digest:
1107
raise KnitCorrupt(self.filename,
1109
'\n of reconstructed text does not match'
1111
'\n for version %s' %
1112
(actual_sha, digest, version_id))
1113
text_map[version_id] = text
1114
return text_map, final_content
1116
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1118
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1119
if version_ids is None:
1120
version_ids = self.versions()
1122
pb = progress.DummyProgress()
1123
# we don't care about inclusions, the caller cares.
1124
# but we need to setup a list of records to visit.
1125
# we need version_id, position, length
1126
version_id_records = []
1127
requested_versions = set(version_ids)
1128
# filter for available versions
1129
for version_id in requested_versions:
1130
if not self.has_version(version_id):
1131
raise RevisionNotPresent(version_id, self.filename)
1132
# get a in-component-order queue:
1133
for version_id in self.versions():
1134
if version_id in requested_versions:
1135
index_memo = self._index.get_position(version_id)
1136
version_id_records.append((version_id, index_memo))
1138
total = len(version_id_records)
1139
for version_idx, (version_id, data, sha_value) in \
1140
enumerate(self._data.read_records_iter(version_id_records)):
1141
pb.update('Walking content.', version_idx, total)
1142
method = self._index.get_method(version_id)
1144
assert method in ('fulltext', 'line-delta')
1145
if method == 'fulltext':
1146
line_iterator = self.factory.get_fulltext_content(data)
1148
line_iterator = self.factory.get_linedelta_content(data)
1149
# XXX: It might be more efficient to yield (version_id,
1150
# line_iterator) in the future. However for now, this is a simpler
1151
# change to integrate into the rest of the codebase. RBC 20071110
1152
for line in line_iterator:
1153
yield line, version_id
1155
pb.update('Walking content.', total, total)
1157
def iter_parents(self, version_ids):
1158
"""Iterate through the parents for many version ids.
1160
:param version_ids: An iterable yielding version_ids.
1161
:return: An iterator that yields (version_id, parents). Requested
1162
version_ids not present in the versioned file are simply skipped.
1163
The order is undefined, allowing for different optimisations in
1164
the underlying implementation.
1166
return self._index.iter_parents(version_ids)
1168
def num_versions(self):
1169
"""See VersionedFile.num_versions()."""
1170
return self._index.num_versions()
1172
__len__ = num_versions
1174
def annotate_iter(self, version_id):
1175
"""See VersionedFile.annotate_iter."""
1176
return self.factory.annotate_iter(self, version_id)
1178
def get_parents(self, version_id):
1179
"""See VersionedFile.get_parents."""
1182
# 52554 calls in 1264 872 internal down from 3674
1184
return self._index.get_parents(version_id)
1186
raise RevisionNotPresent(version_id, self.filename)
1188
def get_parents_with_ghosts(self, version_id):
1189
"""See VersionedFile.get_parents."""
1191
return self._index.get_parents_with_ghosts(version_id)
1193
raise RevisionNotPresent(version_id, self.filename)
1195
def get_ancestry(self, versions, topo_sorted=True):
1196
"""See VersionedFile.get_ancestry."""
1197
if isinstance(versions, basestring):
1198
versions = [versions]
1201
return self._index.get_ancestry(versions, topo_sorted)
1203
def get_ancestry_with_ghosts(self, versions):
1204
"""See VersionedFile.get_ancestry_with_ghosts."""
1205
if isinstance(versions, basestring):
1206
versions = [versions]
1209
return self._index.get_ancestry_with_ghosts(versions)
1211
def plan_merge(self, ver_a, ver_b):
1212
"""See VersionedFile.plan_merge."""
1213
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1214
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1215
annotated_a = self.annotate(ver_a)
1216
annotated_b = self.annotate(ver_b)
1217
return merge._plan_annotate_merge(annotated_a, annotated_b,
1218
ancestors_a, ancestors_b)
1221
class _KnitComponentFile(object):
1222
"""One of the files used to implement a knit database"""
1224
def __init__(self, transport, filename, mode, file_mode=None,
1225
create_parent_dir=False, dir_mode=None):
1226
self._transport = transport
1227
self._filename = filename
1229
self._file_mode = file_mode
1230
self._dir_mode = dir_mode
1231
self._create_parent_dir = create_parent_dir
1232
self._need_to_create = False
1234
def _full_path(self):
1235
"""Return the full path to this file."""
1236
return self._transport.base + self._filename
1238
def check_header(self, fp):
1239
line = fp.readline()
1241
# An empty file can actually be treated as though the file doesn't
1243
raise errors.NoSuchFile(self._full_path())
1244
if line != self.HEADER:
1245
raise KnitHeaderError(badline=line,
1246
filename=self._transport.abspath(self._filename))
1249
return '%s(%s)' % (self.__class__.__name__, self._filename)
1252
class _KnitIndex(_KnitComponentFile):
1253
"""Manages knit index file.
1255
The index is already kept in memory and read on startup, to enable
1256
fast lookups of revision information. The cursor of the index
1257
file is always pointing to the end, making it easy to append
1260
_cache is a cache for fast mapping from version id to a Index
1263
_history is a cache for fast mapping from indexes to version ids.
1265
The index data format is dictionary compressed when it comes to
1266
parent references; a index entry may only have parents that with a
1267
lover index number. As a result, the index is topological sorted.
1269
Duplicate entries may be written to the index for a single version id
1270
if this is done then the latter one completely replaces the former:
1271
this allows updates to correct version and parent information.
1272
Note that the two entries may share the delta, and that successive
1273
annotations and references MUST point to the first entry.
1275
The index file on disc contains a header, followed by one line per knit
1276
record. The same revision can be present in an index file more than once.
1277
The first occurrence gets assigned a sequence number starting from 0.
1279
The format of a single line is
1280
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1281
REVISION_ID is a utf8-encoded revision id
1282
FLAGS is a comma separated list of flags about the record. Values include
1283
no-eol, line-delta, fulltext.
1284
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1285
that the the compressed data starts at.
1286
LENGTH is the ascii representation of the length of the data file.
1287
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1289
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1290
revision id already in the knit that is a parent of REVISION_ID.
1291
The ' :' marker is the end of record marker.
1294
when a write is interrupted to the index file, it will result in a line
1295
that does not end in ' :'. If the ' :' is not present at the end of a line,
1296
or at the end of the file, then the record that is missing it will be
1297
ignored by the parser.
1299
When writing new records to the index file, the data is preceded by '\n'
1300
to ensure that records always start on new lines even if the last write was
1301
interrupted. As a result its normal for the last line in the index to be
1302
missing a trailing newline. One can be added with no harmful effects.
1305
HEADER = "# bzr knit index 8\n"
1307
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1308
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1310
def _cache_version(self, version_id, options, pos, size, parents):
1311
"""Cache a version record in the history array and index cache.
1313
This is inlined into _load_data for performance. KEEP IN SYNC.
1314
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1317
# only want the _history index to reference the 1st index entry
1319
if version_id not in self._cache:
1320
index = len(self._history)
1321
self._history.append(version_id)
1323
index = self._cache[version_id][5]
1324
self._cache[version_id] = (version_id,
1331
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1332
create_parent_dir=False, delay_create=False, dir_mode=None):
1333
_KnitComponentFile.__init__(self, transport, filename, mode,
1334
file_mode=file_mode,
1335
create_parent_dir=create_parent_dir,
1338
# position in _history is the 'official' index for a revision
1339
# but the values may have come from a newer entry.
1340
# so - wc -l of a knit index is != the number of unique names
1344
fp = self._transport.get(self._filename)
1346
# _load_data may raise NoSuchFile if the target knit is
1348
_load_data(self, fp)
1352
if mode != 'w' or not create:
1355
self._need_to_create = True
1357
self._transport.put_bytes_non_atomic(
1358
self._filename, self.HEADER, mode=self._file_mode)
1360
def get_graph(self):
1361
"""Return a list of the node:parents lists from this knit index."""
1362
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1364
def get_ancestry(self, versions, topo_sorted=True):
1365
"""See VersionedFile.get_ancestry."""
1366
# get a graph of all the mentioned versions:
1368
pending = set(versions)
1371
version = pending.pop()
1374
parents = [p for p in cache[version][4] if p in cache]
1376
raise RevisionNotPresent(version, self._filename)
1377
# if not completed and not a ghost
1378
pending.update([p for p in parents if p not in graph])
1379
graph[version] = parents
1382
return topo_sort(graph.items())
1384
def get_ancestry_with_ghosts(self, versions):
1385
"""See VersionedFile.get_ancestry_with_ghosts."""
1386
# get a graph of all the mentioned versions:
1387
self.check_versions_present(versions)
1390
pending = set(versions)
1392
version = pending.pop()
1394
parents = cache[version][4]
1400
pending.update([p for p in parents if p not in graph])
1401
graph[version] = parents
1402
return topo_sort(graph.items())
1404
def iter_parents(self, version_ids):
1405
"""Iterate through the parents for many version ids.
1407
:param version_ids: An iterable yielding version_ids.
1408
:return: An iterator that yields (version_id, parents). Requested
1409
version_ids not present in the versioned file are simply skipped.
1410
The order is undefined, allowing for different optimisations in
1411
the underlying implementation.
1413
for version_id in version_ids:
1415
yield version_id, tuple(self.get_parents(version_id))
1419
def num_versions(self):
1420
return len(self._history)
1422
__len__ = num_versions
1424
def get_versions(self):
1425
"""Get all the versions in the file. not topologically sorted."""
1426
return self._history
1428
def _version_list_to_index(self, versions):
1431
for version in versions:
1432
if version in cache:
1433
# -- inlined lookup() --
1434
result_list.append(str(cache[version][5]))
1435
# -- end lookup () --
1437
result_list.append('.' + version)
1438
return ' '.join(result_list)
1440
def add_version(self, version_id, options, index_memo, parents):
1441
"""Add a version record to the index."""
1442
self.add_versions(((version_id, options, index_memo, parents),))
1444
def add_versions(self, versions, random_id=False):
1445
"""Add multiple versions to the index.
1447
:param versions: a list of tuples:
1448
(version_id, options, pos, size, parents).
1449
:param random_id: If True the ids being added were randomly generated
1450
and no check for existence will be performed.
1453
orig_history = self._history[:]
1454
orig_cache = self._cache.copy()
1457
for version_id, options, (index, pos, size), parents in versions:
1458
line = "\n%s %s %s %s %s :" % (version_id,
1462
self._version_list_to_index(parents))
1463
assert isinstance(line, str), \
1464
'content must be utf-8 encoded: %r' % (line,)
1466
self._cache_version(version_id, options, pos, size, parents)
1467
if not self._need_to_create:
1468
self._transport.append_bytes(self._filename, ''.join(lines))
1471
sio.write(self.HEADER)
1472
sio.writelines(lines)
1474
self._transport.put_file_non_atomic(self._filename, sio,
1475
create_parent_dir=self._create_parent_dir,
1476
mode=self._file_mode,
1477
dir_mode=self._dir_mode)
1478
self._need_to_create = False
1480
# If any problems happen, restore the original values and re-raise
1481
self._history = orig_history
1482
self._cache = orig_cache
1485
def has_version(self, version_id):
1486
"""True if the version is in the index."""
1487
return version_id in self._cache
1489
def get_position(self, version_id):
1490
"""Return details needed to access the version.
1492
.kndx indices do not support split-out data, so return None for the
1495
:return: a tuple (None, data position, size) to hand to the access
1496
logic to get the record.
1498
entry = self._cache[version_id]
1499
return None, entry[2], entry[3]
1501
def get_method(self, version_id):
1502
"""Return compression method of specified version."""
1504
options = self._cache[version_id][1]
1506
raise RevisionNotPresent(version_id, self._filename)
1507
if 'fulltext' in options:
1510
if 'line-delta' not in options:
1511
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1514
def get_options(self, version_id):
1515
"""Return a string represention options.
1519
return self._cache[version_id][1]
1521
def get_parents(self, version_id):
1522
"""Return parents of specified version ignoring ghosts."""
1523
return [parent for parent in self._cache[version_id][4]
1524
if parent in self._cache]
1526
def get_parents_with_ghosts(self, version_id):
1527
"""Return parents of specified version with ghosts."""
1528
return self._cache[version_id][4]
1530
def check_versions_present(self, version_ids):
1531
"""Check that all specified versions are present."""
1533
for version_id in version_ids:
1534
if version_id not in cache:
1535
raise RevisionNotPresent(version_id, self._filename)
1538
class KnitGraphIndex(object):
1539
"""A knit index that builds on GraphIndex."""
1541
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1542
"""Construct a KnitGraphIndex on a graph_index.
1544
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1545
:param deltas: Allow delta-compressed records.
1546
:param add_callback: If not None, allow additions to the index and call
1547
this callback with a list of added GraphIndex nodes:
1548
[(node, value, node_refs), ...]
1549
:param parents: If True, record knits parents, if not do not record
1552
self._graph_index = graph_index
1553
self._deltas = deltas
1554
self._add_callback = add_callback
1555
self._parents = parents
1556
if deltas and not parents:
1557
raise KnitCorrupt(self, "Cannot do delta compression without "
1560
def _get_entries(self, keys, check_present=False):
1561
"""Get the entries for keys.
1563
:param keys: An iterable of index keys, - 1-tuples.
1568
for node in self._graph_index.iter_entries(keys):
1570
found_keys.add(node[1])
1572
# adapt parentless index to the rest of the code.
1573
for node in self._graph_index.iter_entries(keys):
1574
yield node[0], node[1], node[2], ()
1575
found_keys.add(node[1])
1577
missing_keys = keys.difference(found_keys)
1579
raise RevisionNotPresent(missing_keys.pop(), self)
1581
def _present_keys(self, version_ids):
1583
node[1] for node in self._get_entries(version_ids)])
1585
def _parentless_ancestry(self, versions):
1586
"""Honour the get_ancestry API for parentless knit indices."""
1587
wanted_keys = self._version_ids_to_keys(versions)
1588
present_keys = self._present_keys(wanted_keys)
1589
missing = set(wanted_keys).difference(present_keys)
1591
raise RevisionNotPresent(missing.pop(), self)
1592
return list(self._keys_to_version_ids(present_keys))
1594
def get_ancestry(self, versions, topo_sorted=True):
1595
"""See VersionedFile.get_ancestry."""
1596
if not self._parents:
1597
return self._parentless_ancestry(versions)
1598
# XXX: This will do len(history) index calls - perhaps
1599
# it should be altered to be a index core feature?
1600
# get a graph of all the mentioned versions:
1603
versions = self._version_ids_to_keys(versions)
1604
pending = set(versions)
1606
# get all pending nodes
1607
this_iteration = pending
1608
new_nodes = self._get_entries(this_iteration)
1611
for (index, key, value, node_refs) in new_nodes:
1612
# dont ask for ghosties - otherwise
1613
# we we can end up looping with pending
1614
# being entirely ghosted.
1615
graph[key] = [parent for parent in node_refs[0]
1616
if parent not in ghosts]
1618
for parent in graph[key]:
1619
# dont examine known nodes again
1624
ghosts.update(this_iteration.difference(found))
1625
if versions.difference(graph):
1626
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1628
result_keys = topo_sort(graph.items())
1630
result_keys = graph.iterkeys()
1631
return [key[0] for key in result_keys]
1633
def get_ancestry_with_ghosts(self, versions):
1634
"""See VersionedFile.get_ancestry."""
1635
if not self._parents:
1636
return self._parentless_ancestry(versions)
1637
# XXX: This will do len(history) index calls - perhaps
1638
# it should be altered to be a index core feature?
1639
# get a graph of all the mentioned versions:
1641
versions = self._version_ids_to_keys(versions)
1642
pending = set(versions)
1644
# get all pending nodes
1645
this_iteration = pending
1646
new_nodes = self._get_entries(this_iteration)
1648
for (index, key, value, node_refs) in new_nodes:
1649
graph[key] = node_refs[0]
1651
for parent in graph[key]:
1652
# dont examine known nodes again
1656
missing_versions = this_iteration.difference(graph)
1657
missing_needed = versions.intersection(missing_versions)
1659
raise RevisionNotPresent(missing_needed.pop(), self)
1660
for missing_version in missing_versions:
1661
# add a key, no parents
1662
graph[missing_version] = []
1663
pending.discard(missing_version) # don't look for it
1664
result_keys = topo_sort(graph.items())
1665
return [key[0] for key in result_keys]
1667
def get_graph(self):
1668
"""Return a list of the node:parents lists from this knit index."""
1669
if not self._parents:
1670
return [(key, ()) for key in self.get_versions()]
1672
for index, key, value, refs in self._graph_index.iter_all_entries():
1673
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1676
def iter_parents(self, version_ids):
1677
"""Iterate through the parents for many version ids.
1679
:param version_ids: An iterable yielding version_ids.
1680
:return: An iterator that yields (version_id, parents). Requested
1681
version_ids not present in the versioned file are simply skipped.
1682
The order is undefined, allowing for different optimisations in
1683
the underlying implementation.
1686
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1688
present_parents = set()
1689
for node in all_nodes:
1690
all_parents.update(node[3][0])
1691
# any node we are querying must be present
1692
present_parents.add(node[1])
1693
unknown_parents = all_parents.difference(present_parents)
1694
present_parents.update(self._present_keys(unknown_parents))
1695
for node in all_nodes:
1697
for parent in node[3][0]:
1698
if parent in present_parents:
1699
parents.append(parent[0])
1700
yield node[1][0], tuple(parents)
1702
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1703
yield node[1][0], ()
1705
def num_versions(self):
1706
return len(list(self._graph_index.iter_all_entries()))
1708
__len__ = num_versions
1710
def get_versions(self):
1711
"""Get all the versions in the file. not topologically sorted."""
1712
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1714
def has_version(self, version_id):
1715
"""True if the version is in the index."""
1716
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1718
def _keys_to_version_ids(self, keys):
1719
return tuple(key[0] for key in keys)
1721
def get_position(self, version_id):
1722
"""Return details needed to access the version.
1724
:return: a tuple (index, data position, size) to hand to the access
1725
logic to get the record.
1727
node = self._get_node(version_id)
1728
bits = node[2][1:].split(' ')
1729
return node[0], int(bits[0]), int(bits[1])
1731
def get_method(self, version_id):
1732
"""Return compression method of specified version."""
1733
if not self._deltas:
1735
return self._parent_compression(self._get_node(version_id)[3][1])
1737
def _parent_compression(self, reference_list):
1738
# use the second reference list to decide if this is delta'd or not.
1739
if len(reference_list):
1744
def _get_node(self, version_id):
1746
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1748
raise RevisionNotPresent(version_id, self)
1750
def get_options(self, version_id):
1751
"""Return a string represention options.
1755
node = self._get_node(version_id)
1756
if not self._deltas:
1757
options = ['fulltext']
1759
options = [self._parent_compression(node[3][1])]
1760
if node[2][0] == 'N':
1761
options.append('no-eol')
1764
def get_parents(self, version_id):
1765
"""Return parents of specified version ignoring ghosts."""
1766
parents = list(self.iter_parents([version_id]))
1769
raise errors.RevisionNotPresent(version_id, self)
1770
return parents[0][1]
1772
def get_parents_with_ghosts(self, version_id):
1773
"""Return parents of specified version with ghosts."""
1774
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1775
check_present=True))
1776
if not self._parents:
1778
return self._keys_to_version_ids(nodes[0][3][0])
1780
def check_versions_present(self, version_ids):
1781
"""Check that all specified versions are present."""
1782
keys = self._version_ids_to_keys(version_ids)
1783
present = self._present_keys(keys)
1784
missing = keys.difference(present)
1786
raise RevisionNotPresent(missing.pop(), self)
1788
def add_version(self, version_id, options, access_memo, parents):
1789
"""Add a version record to the index."""
1790
return self.add_versions(((version_id, options, access_memo, parents),))
1792
def add_versions(self, versions, random_id=False):
1793
"""Add multiple versions to the index.
1795
This function does not insert data into the Immutable GraphIndex
1796
backing the KnitGraphIndex, instead it prepares data for insertion by
1797
the caller and checks that it is safe to insert then calls
1798
self._add_callback with the prepared GraphIndex nodes.
1800
:param versions: a list of tuples:
1801
(version_id, options, pos, size, parents).
1802
:param random_id: If True the ids being added were randomly generated
1803
and no check for existence will be performed.
1805
if not self._add_callback:
1806
raise errors.ReadOnlyError(self)
1807
# we hope there are no repositories with inconsistent parentage
1812
for (version_id, options, access_memo, parents) in versions:
1813
index, pos, size = access_memo
1814
key = (version_id, )
1815
parents = tuple((parent, ) for parent in parents)
1816
if 'no-eol' in options:
1820
value += "%d %d" % (pos, size)
1821
if not self._deltas:
1822
if 'line-delta' in options:
1823
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1826
if 'line-delta' in options:
1827
node_refs = (parents, (parents[0],))
1829
node_refs = (parents, ())
1831
node_refs = (parents, )
1834
raise KnitCorrupt(self, "attempt to add node with parents "
1835
"in parentless index.")
1837
keys[key] = (value, node_refs)
1839
present_nodes = self._get_entries(keys)
1840
for (index, key, value, node_refs) in present_nodes:
1841
if (value, node_refs) != keys[key]:
1842
raise KnitCorrupt(self, "inconsistent details in add_versions"
1843
": %s %s" % ((value, node_refs), keys[key]))
1847
for key, (value, node_refs) in keys.iteritems():
1848
result.append((key, value, node_refs))
1850
for key, (value, node_refs) in keys.iteritems():
1851
result.append((key, value))
1852
self._add_callback(result)
1854
def _version_ids_to_keys(self, version_ids):
1855
return set((version_id, ) for version_id in version_ids)
1858
class _KnitAccess(object):
1859
"""Access to knit records in a .knit file."""
1861
def __init__(self, transport, filename, _file_mode, _dir_mode,
1862
_need_to_create, _create_parent_dir):
1863
"""Create a _KnitAccess for accessing and inserting data.
1865
:param transport: The transport the .knit is located on.
1866
:param filename: The filename of the .knit.
1868
self._transport = transport
1869
self._filename = filename
1870
self._file_mode = _file_mode
1871
self._dir_mode = _dir_mode
1872
self._need_to_create = _need_to_create
1873
self._create_parent_dir = _create_parent_dir
1875
def add_raw_records(self, sizes, raw_data):
1876
"""Add raw knit bytes to a storage area.
1878
The data is spooled to whereever the access method is storing data.
1880
:param sizes: An iterable containing the size of each raw data segment.
1881
:param raw_data: A bytestring containing the data.
1882
:return: A list of memos to retrieve the record later. Each memo is a
1883
tuple - (index, pos, length), where the index field is always None
1884
for the .knit access method.
1886
assert type(raw_data) == str, \
1887
'data must be plain bytes was %s' % type(raw_data)
1888
if not self._need_to_create:
1889
base = self._transport.append_bytes(self._filename, raw_data)
1891
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1892
create_parent_dir=self._create_parent_dir,
1893
mode=self._file_mode,
1894
dir_mode=self._dir_mode)
1895
self._need_to_create = False
1899
result.append((None, base, size))
1904
"""IFF this data access has its own storage area, initialise it.
1908
self._transport.put_bytes_non_atomic(self._filename, '',
1909
mode=self._file_mode)
1911
def open_file(self):
1912
"""IFF this data access can be represented as a single file, open it.
1914
For knits that are not mapped to a single file on disk this will
1917
:return: None or a file handle.
1920
return self._transport.get(self._filename)
1925
def get_raw_records(self, memos_for_retrieval):
1926
"""Get the raw bytes for a records.
1928
:param memos_for_retrieval: An iterable containing the (index, pos,
1929
length) memo for retrieving the bytes. The .knit method ignores
1930
the index as there is always only a single file.
1931
:return: An iterator over the bytes of the records.
1933
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1934
for pos, data in self._transport.readv(self._filename, read_vector):
1938
class _PackAccess(object):
1939
"""Access to knit records via a collection of packs."""
1941
def __init__(self, index_to_packs, writer=None):
1942
"""Create a _PackAccess object.
1944
:param index_to_packs: A dict mapping index objects to the transport
1945
and file names for obtaining data.
1946
:param writer: A tuple (pack.ContainerWriter, write_index) which
1947
contains the pack to write, and the index that reads from it will
1951
self.container_writer = writer[0]
1952
self.write_index = writer[1]
1954
self.container_writer = None
1955
self.write_index = None
1956
self.indices = index_to_packs
1958
def add_raw_records(self, sizes, raw_data):
1959
"""Add raw knit bytes to a storage area.
1961
The data is spooled to the container writer in one bytes-record per
1964
:param sizes: An iterable containing the size of each raw data segment.
1965
:param raw_data: A bytestring containing the data.
1966
:return: A list of memos to retrieve the record later. Each memo is a
1967
tuple - (index, pos, length), where the index field is the
1968
write_index object supplied to the PackAccess object.
1970
assert type(raw_data) == str, \
1971
'data must be plain bytes was %s' % type(raw_data)
1975
p_offset, p_length = self.container_writer.add_bytes_record(
1976
raw_data[offset:offset+size], [])
1978
result.append((self.write_index, p_offset, p_length))
1982
"""Pack based knits do not get individually created."""
1984
def get_raw_records(self, memos_for_retrieval):
1985
"""Get the raw bytes for a records.
1987
:param memos_for_retrieval: An iterable containing the (index, pos,
1988
length) memo for retrieving the bytes. The Pack access method
1989
looks up the pack to use for a given record in its index_to_pack
1991
:return: An iterator over the bytes of the records.
1993
# first pass, group into same-index requests
1995
current_index = None
1996
for (index, offset, length) in memos_for_retrieval:
1997
if current_index == index:
1998
current_list.append((offset, length))
2000
if current_index is not None:
2001
request_lists.append((current_index, current_list))
2002
current_index = index
2003
current_list = [(offset, length)]
2004
# handle the last entry
2005
if current_index is not None:
2006
request_lists.append((current_index, current_list))
2007
for index, offsets in request_lists:
2008
transport, path = self.indices[index]
2009
reader = pack.make_readv_reader(transport, path, offsets)
2010
for names, read_func in reader.iter_records():
2011
yield read_func(None)
2013
def open_file(self):
2014
"""Pack based knits have no single file."""
2017
def set_writer(self, writer, index, (transport, packname)):
2018
"""Set a writer to use for adding data."""
2019
if index is not None:
2020
self.indices[index] = (transport, packname)
2021
self.container_writer = writer
2022
self.write_index = index
2025
class _KnitData(object):
2026
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2028
The KnitData class provides the logic for parsing and using knit records,
2029
making use of an access method for the low level read and write operations.
2032
def __init__(self, access):
2033
"""Create a KnitData object.
2035
:param access: The access method to use. Access methods such as
2036
_KnitAccess manage the insertion of raw records and the subsequent
2037
retrieval of the same.
2039
self._access = access
2040
self._checked = False
2041
# TODO: jam 20060713 conceptually, this could spill to disk
2042
# if the cached size gets larger than a certain amount
2043
# but it complicates the model a bit, so for now just use
2044
# a simple dictionary
2046
self._do_cache = False
2048
def enable_cache(self):
2049
"""Enable caching of reads."""
2050
self._do_cache = True
2052
def clear_cache(self):
2053
"""Clear the record cache."""
2054
self._do_cache = False
2057
def _open_file(self):
2058
return self._access.open_file()
2060
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2061
"""Convert version_id, digest, lines into a raw data block.
2063
:param dense_lines: The bytes of lines but in a denser form. For
2064
instance, if lines is a list of 1000 bytestrings each ending in \n,
2065
dense_lines may be a list with one line in it, containing all the
2066
1000's lines and their \n's. Using dense_lines if it is already
2067
known is a win because the string join to create bytes in this
2068
function spends less time resizing the final string.
2069
:return: (len, a StringIO instance with the raw data ready to read.)
2071
# Note: using a string copy here increases memory pressure with e.g.
2072
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2073
# when doing the initial commit of a mozilla tree. RBC 20070921
2074
bytes = ''.join(chain(
2075
["version %s %d %s\n" % (version_id,
2078
dense_lines or lines,
2079
["end %s\n" % version_id]))
2080
assert bytes.__class__ == str
2081
compressed_bytes = bytes_to_gzip(bytes)
2082
return len(compressed_bytes), compressed_bytes
2084
def add_raw_records(self, sizes, raw_data):
2085
"""Append a prepared record to the data file.
2087
:param sizes: An iterable containing the size of each raw data segment.
2088
:param raw_data: A bytestring containing the data.
2089
:return: a list of index data for the way the data was stored.
2090
See the access method add_raw_records documentation for more
2093
return self._access.add_raw_records(sizes, raw_data)
2095
def _parse_record_header(self, version_id, raw_data):
2096
"""Parse a record header for consistency.
2098
:return: the header and the decompressor stream.
2099
as (stream, header_record)
2101
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2103
rec = self._check_header(version_id, df.readline())
2104
except Exception, e:
2105
raise KnitCorrupt(self._access,
2106
"While reading {%s} got %s(%s)"
2107
% (version_id, e.__class__.__name__, str(e)))
2110
def _check_header(self, version_id, line):
2113
raise KnitCorrupt(self._access,
2114
'unexpected number of elements in record header')
2115
if rec[1] != version_id:
2116
raise KnitCorrupt(self._access,
2117
'unexpected version, wanted %r, got %r'
2118
% (version_id, rec[1]))
2121
def _parse_record(self, version_id, data):
2123
# 4168 calls in 2880 217 internal
2124
# 4168 calls to _parse_record_header in 2121
2125
# 4168 calls to readlines in 330
2126
df = GzipFile(mode='rb', fileobj=StringIO(data))
2129
record_contents = df.readlines()
2130
except Exception, e:
2131
raise KnitCorrupt(self._access,
2132
"While reading {%s} got %s(%s)"
2133
% (version_id, e.__class__.__name__, str(e)))
2134
header = record_contents.pop(0)
2135
rec = self._check_header(version_id, header)
2137
last_line = record_contents.pop()
2138
if len(record_contents) != int(rec[2]):
2139
raise KnitCorrupt(self._access,
2140
'incorrect number of lines %s != %s'
2142
% (len(record_contents), int(rec[2]),
2144
if last_line != 'end %s\n' % rec[1]:
2145
raise KnitCorrupt(self._access,
2146
'unexpected version end line %r, wanted %r'
2147
% (last_line, version_id))
2149
return record_contents, rec[3]
2151
def read_records_iter_raw(self, records):
2152
"""Read text records from data file and yield raw data.
2154
This unpacks enough of the text record to validate the id is
2155
as expected but thats all.
2157
# setup an iterator of the external records:
2158
# uses readv so nice and fast we hope.
2160
# grab the disk data needed.
2162
# Don't check _cache if it is empty
2163
needed_offsets = [index_memo for version_id, index_memo
2165
if version_id not in self._cache]
2167
needed_offsets = [index_memo for version_id, index_memo
2170
raw_records = self._access.get_raw_records(needed_offsets)
2172
for version_id, index_memo in records:
2173
if version_id in self._cache:
2174
# This data has already been validated
2175
data = self._cache[version_id]
2177
data = raw_records.next()
2179
self._cache[version_id] = data
2181
# validate the header
2182
df, rec = self._parse_record_header(version_id, data)
2184
yield version_id, data
2186
def read_records_iter(self, records):
2187
"""Read text records from data file and yield result.
2189
The result will be returned in whatever is the fastest to read.
2190
Not by the order requested. Also, multiple requests for the same
2191
record will only yield 1 response.
2192
:param records: A list of (version_id, pos, len) entries
2193
:return: Yields (version_id, contents, digest) in the order
2194
read, not the order requested
2200
# Skip records we have alread seen
2201
yielded_records = set()
2202
needed_records = set()
2203
for record in records:
2204
if record[0] in self._cache:
2205
if record[0] in yielded_records:
2207
yielded_records.add(record[0])
2208
data = self._cache[record[0]]
2209
content, digest = self._parse_record(record[0], data)
2210
yield (record[0], content, digest)
2212
needed_records.add(record)
2213
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2215
needed_records = sorted(set(records), key=operator.itemgetter(1))
2217
if not needed_records:
2220
# The transport optimizes the fetching as well
2221
# (ie, reads continuous ranges.)
2222
raw_data = self._access.get_raw_records(
2223
[index_memo for version_id, index_memo in needed_records])
2225
for (version_id, index_memo), data in \
2226
izip(iter(needed_records), raw_data):
2227
content, digest = self._parse_record(version_id, data)
2229
self._cache[version_id] = data
2230
yield version_id, content, digest
2232
def read_records(self, records):
2233
"""Read records into a dictionary."""
2235
for record_id, content, digest in \
2236
self.read_records_iter(records):
2237
components[record_id] = (content, digest)
2241
class InterKnit(InterVersionedFile):
2242
"""Optimised code paths for knit to knit operations."""
2244
_matching_file_from_factory = KnitVersionedFile
2245
_matching_file_to_factory = KnitVersionedFile
2248
def is_compatible(source, target):
2249
"""Be compatible with knits. """
2251
return (isinstance(source, KnitVersionedFile) and
2252
isinstance(target, KnitVersionedFile))
2253
except AttributeError:
2256
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2257
"""Copy texts to the target by extracting and adding them one by one.
2259
see join() for the parameter definitions.
2261
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2262
graph = self.source.get_graph(version_ids)
2263
order = topo_sort(graph.items())
2265
def size_of_content(content):
2266
return sum(len(line) for line in content.text())
2267
# Cache at most 10MB of parent texts
2268
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2269
compute_size=size_of_content)
2270
# TODO: jam 20071116 It would be nice to have a streaming interface to
2271
# get multiple texts from a source. The source could be smarter
2272
# about how it handled intermediate stages.
2273
# get_line_list() or make_mpdiffs() seem like a possibility, but
2274
# at the moment they extract all full texts into memory, which
2275
# causes us to store more than our 3x fulltext goal.
2276
# Repository.iter_files_bytes() may be another possibility
2277
to_process = [version for version in order
2278
if version not in self.target]
2279
total = len(to_process)
2280
pb = ui.ui_factory.nested_progress_bar()
2282
for index, version in enumerate(to_process):
2283
pb.update('Converting versioned data', index, total)
2284
sha1, num_bytes, parent_text = self.target.add_lines(version,
2285
self.source.get_parents(version),
2286
self.source.get_lines(version),
2287
parent_texts=parent_cache)
2288
parent_cache[version] = parent_text
2293
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2294
"""See InterVersionedFile.join."""
2295
assert isinstance(self.source, KnitVersionedFile)
2296
assert isinstance(self.target, KnitVersionedFile)
2298
# If the source and target are mismatched w.r.t. annotations vs
2299
# plain, the data needs to be converted accordingly
2300
if self.source.factory.annotated == self.target.factory.annotated:
2302
elif self.source.factory.annotated:
2303
converter = self._anno_to_plain_converter
2305
# We're converting from a plain to an annotated knit. Copy them
2306
# across by full texts.
2307
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2309
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2313
pb = ui.ui_factory.nested_progress_bar()
2315
version_ids = list(version_ids)
2316
if None in version_ids:
2317
version_ids.remove(None)
2319
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2320
this_versions = set(self.target._index.get_versions())
2321
# XXX: For efficiency we should not look at the whole index,
2322
# we only need to consider the referenced revisions - they
2323
# must all be present, or the method must be full-text.
2324
# TODO, RBC 20070919
2325
needed_versions = self.source_ancestry - this_versions
2327
if not needed_versions:
2329
full_list = topo_sort(self.source.get_graph())
2331
version_list = [i for i in full_list if (not self.target.has_version(i)
2332
and i in needed_versions)]
2336
copy_queue_records = []
2338
for version_id in version_list:
2339
options = self.source._index.get_options(version_id)
2340
parents = self.source._index.get_parents_with_ghosts(version_id)
2341
# check that its will be a consistent copy:
2342
for parent in parents:
2343
# if source has the parent, we must :
2344
# * already have it or
2345
# * have it scheduled already
2346
# otherwise we don't care
2347
assert (self.target.has_version(parent) or
2348
parent in copy_set or
2349
not self.source.has_version(parent))
2350
index_memo = self.source._index.get_position(version_id)
2351
copy_queue_records.append((version_id, index_memo))
2352
copy_queue.append((version_id, options, parents))
2353
copy_set.add(version_id)
2355
# data suck the join:
2357
total = len(version_list)
2360
for (version_id, raw_data), \
2361
(version_id2, options, parents) in \
2362
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2364
assert version_id == version_id2, 'logic error, inconsistent results'
2366
pb.update("Joining knit", count, total)
2368
size, raw_data = converter(raw_data, version_id, options,
2371
size = len(raw_data)
2372
raw_records.append((version_id, options, parents, size))
2373
raw_datum.append(raw_data)
2374
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2379
def _anno_to_plain_converter(self, raw_data, version_id, options,
2381
"""Convert annotated content to plain content."""
2382
data, digest = self.source._data._parse_record(version_id, raw_data)
2383
if 'fulltext' in options:
2384
content = self.source.factory.parse_fulltext(data, version_id)
2385
lines = self.target.factory.lower_fulltext(content)
2387
delta = self.source.factory.parse_line_delta(data, version_id,
2389
lines = self.target.factory.lower_line_delta(delta)
2390
return self.target._data._record_to_data(version_id, digest, lines)
2393
InterVersionedFile.register_optimiser(InterKnit)
2396
class WeaveToKnit(InterVersionedFile):
2397
"""Optimised code paths for weave to knit operations."""
2399
_matching_file_from_factory = bzrlib.weave.WeaveFile
2400
_matching_file_to_factory = KnitVersionedFile
2403
def is_compatible(source, target):
2404
"""Be compatible with weaves to knits."""
2406
return (isinstance(source, bzrlib.weave.Weave) and
2407
isinstance(target, KnitVersionedFile))
2408
except AttributeError:
2411
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2412
"""See InterVersionedFile.join."""
2413
assert isinstance(self.source, bzrlib.weave.Weave)
2414
assert isinstance(self.target, KnitVersionedFile)
2416
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2421
pb = ui.ui_factory.nested_progress_bar()
2423
version_ids = list(version_ids)
2425
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2426
this_versions = set(self.target._index.get_versions())
2427
needed_versions = self.source_ancestry - this_versions
2429
if not needed_versions:
2431
full_list = topo_sort(self.source.get_graph())
2433
version_list = [i for i in full_list if (not self.target.has_version(i)
2434
and i in needed_versions)]
2438
total = len(version_list)
2439
for version_id in version_list:
2440
pb.update("Converting to knit", count, total)
2441
parents = self.source.get_parents(version_id)
2442
# check that its will be a consistent copy:
2443
for parent in parents:
2444
# if source has the parent, we must already have it
2445
assert (self.target.has_version(parent))
2446
self.target.add_lines(
2447
version_id, parents, self.source.get_lines(version_id))
2454
InterVersionedFile.register_optimiser(WeaveToKnit)
2457
# Deprecated, use PatienceSequenceMatcher instead
2458
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2461
def annotate_knit(knit, revision_id):
2462
"""Annotate a knit with no cached annotations.
2464
This implementation is for knits with no cached annotations.
2465
It will work for knits with cached annotations, but this is not
2468
ancestry = knit.get_ancestry(revision_id)
2469
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2471
for candidate in ancestry:
2472
if candidate in annotations:
2474
parents = knit.get_parents(candidate)
2475
if len(parents) == 0:
2477
elif knit._index.get_method(candidate) != 'line-delta':
2480
parent, sha1, noeol, delta = knit.get_delta(candidate)
2481
blocks = KnitContent.get_line_delta_blocks(delta,
2482
fulltext[parents[0]], fulltext[candidate])
2483
annotations[candidate] = list(annotate.reannotate([annotations[p]
2484
for p in parents], fulltext[candidate], candidate, blocks))
2485
return iter(annotations[revision_id])
2489
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2491
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
