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 (
101
RevisionAlreadyPresent,
103
from bzrlib.tuned_gzip import GzipFile, bytes_to_gzip
104
from bzrlib.osutils import (
110
from bzrlib.symbol_versioning import DEPRECATED_PARAMETER, deprecated_passed
111
from bzrlib.tsort import topo_sort
114
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
117
# TODO: Split out code specific to this format into an associated object.
119
# TODO: Can we put in some kind of value to check that the index and data
120
# files belong together?
122
# TODO: accommodate binaries, perhaps by storing a byte count
124
# TODO: function to check whole file
126
# TODO: atomically append data, then measure backwards from the cursor
127
# position after writing to work out where it was located. we may need to
128
# bypass python file buffering.
130
DATA_SUFFIX = '.knit'
131
INDEX_SUFFIX = '.kndx'
134
class KnitContent(object):
135
"""Content of a knit version to which deltas can be applied."""
138
"""Return a list of (origin, text) tuples."""
139
return list(self.annotate_iter())
141
def apply_delta(self, delta, new_version_id):
142
"""Apply delta to this object to become new_version_id."""
143
raise NotImplementedError(self.apply_delta)
145
def line_delta_iter(self, new_lines):
146
"""Generate line-based delta from this content to new_lines."""
147
new_texts = new_lines.text()
148
old_texts = self.text()
149
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
150
for tag, i1, i2, j1, j2 in s.get_opcodes():
153
# ofrom, oto, length, data
154
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
156
def line_delta(self, new_lines):
157
return list(self.line_delta_iter(new_lines))
160
def get_line_delta_blocks(knit_delta, source, target):
161
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
162
target_len = len(target)
165
for s_begin, s_end, t_len, new_text in knit_delta:
166
true_n = s_begin - s_pos
169
# knit deltas do not provide reliable info about whether the
170
# last line of a file matches, due to eol handling.
171
if source[s_pos + n -1] != target[t_pos + n -1]:
174
yield s_pos, t_pos, n
175
t_pos += t_len + true_n
177
n = target_len - t_pos
179
if source[s_pos + n -1] != target[t_pos + n -1]:
182
yield s_pos, t_pos, n
183
yield s_pos + (target_len - t_pos), target_len, 0
186
class AnnotatedKnitContent(KnitContent):
187
"""Annotated content."""
189
def __init__(self, lines):
192
def annotate_iter(self):
193
"""Yield tuples of (origin, text) for each content line."""
194
return iter(self._lines)
196
def apply_delta(self, delta, new_version_id):
197
"""Apply delta to this object to become new_version_id."""
200
for start, end, count, delta_lines in delta:
201
lines[offset+start:offset+end] = delta_lines
202
offset = offset + (start - end) + count
204
def strip_last_line_newline(self):
205
line = self._lines[-1][1].rstrip('\n')
206
self._lines[-1] = (self._lines[-1][0], line)
210
return [text for origin, text in self._lines]
211
except ValueError, e:
212
# most commonly (only?) caused by the internal form of the knit
213
# missing annotation information because of a bug - see thread
215
raise KnitCorrupt(self,
216
"line in annotated knit missing annotation information: %s"
220
return AnnotatedKnitContent(self._lines[:])
223
class PlainKnitContent(KnitContent):
224
"""Unannotated content.
226
When annotate[_iter] is called on this content, the same version is reported
227
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
231
def __init__(self, lines, version_id):
233
self._version_id = version_id
235
def annotate_iter(self):
236
"""Yield tuples of (origin, text) for each content line."""
237
for line in self._lines:
238
yield self._version_id, line
240
def apply_delta(self, delta, new_version_id):
241
"""Apply delta to this object to become new_version_id."""
244
for start, end, count, delta_lines in delta:
245
lines[offset+start:offset+end] = delta_lines
246
offset = offset + (start - end) + count
247
self._version_id = new_version_id
250
return PlainKnitContent(self._lines[:], self._version_id)
252
def strip_last_line_newline(self):
253
self._lines[-1] = self._lines[-1].rstrip('\n')
259
class KnitAnnotateFactory(object):
260
"""Factory for creating annotated Content objects."""
264
def make(self, lines, version_id):
265
num_lines = len(lines)
266
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
268
def parse_fulltext(self, content, version_id):
269
"""Convert fulltext to internal representation
271
fulltext content is of the format
272
revid(utf8) plaintext\n
273
internal representation is of the format:
276
# TODO: jam 20070209 The tests expect this to be returned as tuples,
277
# but the code itself doesn't really depend on that.
278
# Figure out a way to not require the overhead of turning the
279
# list back into tuples.
280
lines = [tuple(line.split(' ', 1)) for line in content]
281
return AnnotatedKnitContent(lines)
283
def parse_line_delta_iter(self, lines):
284
return iter(self.parse_line_delta(lines))
286
def parse_line_delta(self, lines, version_id, plain=False):
287
"""Convert a line based delta into internal representation.
289
line delta is in the form of:
290
intstart intend intcount
292
revid(utf8) newline\n
293
internal representation is
294
(start, end, count, [1..count tuples (revid, newline)])
296
:param plain: If True, the lines are returned as a plain
297
list without annotations, not as a list of (origin, content) tuples, i.e.
298
(start, end, count, [1..count newline])
305
def cache_and_return(line):
306
origin, text = line.split(' ', 1)
307
return cache.setdefault(origin, origin), text
309
# walk through the lines parsing.
310
# Note that the plain test is explicitly pulled out of the
311
# loop to minimise any performance impact
314
start, end, count = [int(n) for n in header.split(',')]
315
contents = [next().split(' ', 1)[1] for i in xrange(count)]
316
result.append((start, end, count, contents))
319
start, end, count = [int(n) for n in header.split(',')]
320
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
321
result.append((start, end, count, contents))
324
def get_fulltext_content(self, lines):
325
"""Extract just the content lines from a fulltext."""
326
return (line.split(' ', 1)[1] for line in lines)
328
def get_linedelta_content(self, lines):
329
"""Extract just the content from a line delta.
331
This doesn't return all of the extra information stored in a delta.
332
Only the actual content lines.
337
header = header.split(',')
338
count = int(header[2])
339
for i in xrange(count):
340
origin, text = next().split(' ', 1)
343
def lower_fulltext(self, content):
344
"""convert a fulltext content record into a serializable form.
346
see parse_fulltext which this inverts.
348
# TODO: jam 20070209 We only do the caching thing to make sure that
349
# the origin is a valid utf-8 line, eventually we could remove it
350
return ['%s %s' % (o, t) for o, t in content._lines]
352
def lower_line_delta(self, delta):
353
"""convert a delta into a serializable form.
355
See parse_line_delta which this inverts.
357
# TODO: jam 20070209 We only do the caching thing to make sure that
358
# the origin is a valid utf-8 line, eventually we could remove it
360
for start, end, c, lines in delta:
361
out.append('%d,%d,%d\n' % (start, end, c))
362
out.extend(origin + ' ' + text
363
for origin, text in lines)
366
def annotate_iter(self, knit, version_id):
367
content = knit._get_content(version_id)
368
return content.annotate_iter()
371
class KnitPlainFactory(object):
372
"""Factory for creating plain Content objects."""
376
def make(self, lines, version_id):
377
return PlainKnitContent(lines, version_id)
379
def parse_fulltext(self, content, version_id):
380
"""This parses an unannotated fulltext.
382
Note that this is not a noop - the internal representation
383
has (versionid, line) - its just a constant versionid.
385
return self.make(content, version_id)
387
def parse_line_delta_iter(self, lines, version_id):
389
num_lines = len(lines)
390
while cur < num_lines:
393
start, end, c = [int(n) for n in header.split(',')]
394
yield start, end, c, lines[cur:cur+c]
397
def parse_line_delta(self, lines, version_id):
398
return list(self.parse_line_delta_iter(lines, version_id))
400
def get_fulltext_content(self, lines):
401
"""Extract just the content lines from a fulltext."""
404
def get_linedelta_content(self, lines):
405
"""Extract just the content from a line delta.
407
This doesn't return all of the extra information stored in a delta.
408
Only the actual content lines.
413
header = header.split(',')
414
count = int(header[2])
415
for i in xrange(count):
418
def lower_fulltext(self, content):
419
return content.text()
421
def lower_line_delta(self, delta):
423
for start, end, c, lines in delta:
424
out.append('%d,%d,%d\n' % (start, end, c))
428
def annotate_iter(self, knit, version_id):
429
return annotate_knit(knit, version_id)
432
def make_empty_knit(transport, relpath):
433
"""Construct a empty knit at the specified location."""
434
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
437
class KnitVersionedFile(VersionedFile):
438
"""Weave-like structure with faster random access.
440
A knit stores a number of texts and a summary of the relationships
441
between them. Texts are identified by a string version-id. Texts
442
are normally stored and retrieved as a series of lines, but can
443
also be passed as single strings.
445
Lines are stored with the trailing newline (if any) included, to
446
avoid special cases for files with no final newline. Lines are
447
composed of 8-bit characters, not unicode. The combination of
448
these approaches should mean any 'binary' file can be safely
449
stored and retrieved.
452
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
453
factory=None, delta=True, create=False, create_parent_dir=False,
454
delay_create=False, dir_mode=None, index=None, access_method=None):
455
"""Construct a knit at location specified by relpath.
457
:param create: If not True, only open an existing knit.
458
:param create_parent_dir: If True, create the parent directory if
459
creating the file fails. (This is used for stores with
460
hash-prefixes that may not exist yet)
461
:param delay_create: The calling code is aware that the knit won't
462
actually be created until the first data is stored.
463
:param index: An index to use for the knit.
465
if access_mode is None:
467
super(KnitVersionedFile, self).__init__(access_mode)
468
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
469
self.transport = transport
470
self.filename = relpath
471
self.factory = factory or KnitAnnotateFactory()
472
self.writable = (access_mode == 'w')
475
self._max_delta_chain = 200
478
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
479
access_mode, create=create, file_mode=file_mode,
480
create_parent_dir=create_parent_dir, delay_create=delay_create,
484
if access_method is None:
485
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
486
((create and not len(self)) and delay_create), create_parent_dir)
488
_access = access_method
489
if create and not len(self) and not delay_create:
491
self._data = _KnitData(_access)
494
return '%s(%s)' % (self.__class__.__name__,
495
self.transport.abspath(self.filename))
497
def _check_should_delta(self, first_parents):
498
"""Iterate back through the parent listing, looking for a fulltext.
500
This is used when we want to decide whether to add a delta or a new
501
fulltext. It searches for _max_delta_chain parents. When it finds a
502
fulltext parent, it sees if the total size of the deltas leading up to
503
it is large enough to indicate that we want a new full text anyway.
505
Return True if we should create a new delta, False if we should use a
510
delta_parents = first_parents
511
for count in xrange(self._max_delta_chain):
512
parent = delta_parents[0]
513
method = self._index.get_method(parent)
514
index, pos, size = self._index.get_position(parent)
515
if method == 'fulltext':
519
delta_parents = self._index.get_parents(parent)
521
# We couldn't find a fulltext, so we must create a new one
524
return fulltext_size > delta_size
526
def _add_raw_records(self, records, data):
527
"""Add all the records 'records' with data pre-joined in 'data'.
529
:param records: A list of tuples(version_id, options, parents, size).
530
:param data: The data for the records. When it is written, the records
531
are adjusted to have pos pointing into data by the sum of
532
the preceding records sizes.
535
raw_record_sizes = [record[3] for record in records]
536
positions = self._data.add_raw_records(raw_record_sizes, data)
539
for (version_id, options, parents, size), access_memo in zip(
541
index_entries.append((version_id, options, access_memo, parents))
542
if self._data._do_cache:
543
self._data._cache[version_id] = data[offset:offset+size]
545
self._index.add_versions(index_entries)
547
def enable_cache(self):
548
"""Start caching data for this knit"""
549
self._data.enable_cache()
551
def clear_cache(self):
552
"""Clear the data cache only."""
553
self._data.clear_cache()
555
def copy_to(self, name, transport):
556
"""See VersionedFile.copy_to()."""
557
# copy the current index to a temp index to avoid racing with local
559
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
560
self.transport.get(self._index._filename))
562
f = self._data._open_file()
564
transport.put_file(name + DATA_SUFFIX, f)
567
# move the copied index into place
568
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
570
def create_empty(self, name, transport, mode=None):
571
return KnitVersionedFile(name, transport, factory=self.factory,
572
delta=self.delta, create=True)
574
def get_data_stream(self, required_versions):
575
"""Get a data stream for the specified versions.
577
Versions may be returned in any order, not necessarily the order
578
specified. They are returned in a partial order by compression
579
parent, so that the deltas can be applied as the data stream is
580
inserted; however note that compression parents will not be sent
581
unless they were specifically requested, as the client may already
584
:param required_versions: The exact set of versions to be extracted.
585
Unlike some other knit methods, this is not used to generate a
586
transitive closure, rather it is used precisely as given.
588
:returns: format_signature, list of (version, options, length, parents),
591
required_version_set = frozenset(required_versions)
593
# list of revisions that can just be sent without waiting for their
596
# map from revision to the children based on it
598
# first, read all relevant index data, enough to sort into the right
600
for version_id in required_versions:
601
options = self._index.get_options(version_id)
602
parents = self._index.get_parents_with_ghosts(version_id)
603
index_memo = self._index.get_position(version_id)
604
version_index[version_id] = (index_memo, options, parents)
605
if ('line-delta' in options
606
and parents[0] in required_version_set):
607
# must wait until the parent has been sent
608
deferred.setdefault(parents[0], []). \
611
# either a fulltext, or a delta whose parent the client did
612
# not ask for and presumably already has
613
ready_to_send.append(version_id)
614
# build a list of results to return, plus instructions for data to
616
copy_queue_records = []
617
temp_version_list = []
619
# XXX: pushing and popping lists may be a bit inefficient
620
version_id = ready_to_send.pop(0)
621
(index_memo, options, parents) = version_index[version_id]
622
copy_queue_records.append((version_id, index_memo))
623
none, data_pos, data_size = index_memo
624
temp_version_list.append((version_id, options, data_size,
626
if version_id in deferred:
627
# now we can send all the children of this revision - we could
628
# put them in anywhere, but we hope that sending them soon
629
# after the fulltext will give good locality in the receiver
630
ready_to_send[:0] = deferred.pop(version_id)
631
assert len(deferred) == 0, \
632
"Still have compressed child versions waiting to be sent"
633
# XXX: The stream format is such that we cannot stream it - we have to
634
# know the length of all the data a-priori.
636
result_version_list = []
637
for (version_id, raw_data), \
638
(version_id2, options, _, parents) in \
639
izip(self._data.read_records_iter_raw(copy_queue_records),
641
assert version_id == version_id2, \
642
'logic error, inconsistent results'
643
raw_datum.append(raw_data)
644
result_version_list.append(
645
(version_id, options, len(raw_data), parents))
646
# provide a callback to get data incrementally.
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
source = self._knit_from_datastream(
741
(format, data_list, reader_callable))
745
for version_id, options, length, parents in data_list:
746
if self.has_version(version_id):
747
# First check: the list of parents.
748
my_parents = self.get_parents_with_ghosts(version_id)
749
if my_parents != parents:
750
# XXX: KnitCorrupt is not quite the right exception here.
753
'parents list %r from data stream does not match '
754
'already recorded parents %r for %s'
755
% (parents, my_parents, version_id))
757
# Also check the SHA-1 of the fulltext this content will
759
raw_data = reader_callable(length)
760
my_fulltext_sha1 = self.get_sha1(version_id)
761
df, rec = self._data._parse_record_header(version_id, raw_data)
762
stream_fulltext_sha1 = rec[3]
763
if my_fulltext_sha1 != stream_fulltext_sha1:
764
# Actually, we don't know if it's this knit that's corrupt,
765
# or the data stream we're trying to insert.
767
self.filename, 'sha-1 does not match %s' % version_id)
769
if 'line-delta' in options:
770
# Make sure that this knit record is actually useful: a
771
# line-delta is no use unless we have its parent.
772
# Fetching from a broken repository with this problem
773
# shouldn't break the target repository.
775
# See https://bugs.launchpad.net/bzr/+bug/164443
776
if not self._index.has_version(parents[0]):
779
'line-delta from stream '
782
'missing parent %s\n'
783
'Try running "bzr check" '
784
'on the source repository, and "bzr reconcile" '
786
(version_id, parents[0]))
787
self._add_raw_records(
788
[(version_id, options, parents, length)],
789
reader_callable(length))
791
def _knit_from_datastream(self, (format, data_list, reader_callable)):
792
"""Create a knit object from a data stream.
794
This method exists to allow conversion of data streams that do not
795
match the signature of this knit. Generally it will be slower and use
796
more memory to use this method to insert data, but it will work.
798
:seealso: get_data_stream for details on datastreams.
799
:return: A knit versioned file which can be used to join the datastream
802
if format == "knit-plain":
803
factory = KnitPlainFactory()
804
elif format == "knit-annotated":
805
factory = KnitAnnotateFactory()
807
raise errors.KnitDataStreamUnknown(format)
808
index = _StreamIndex(data_list)
809
access = _StreamAccess(reader_callable, index, self, factory)
810
return KnitVersionedFile(self.filename, self.transport,
811
factory=factory, index=index, access_method=access)
814
"""See VersionedFile.versions."""
815
if 'evil' in debug.debug_flags:
816
trace.mutter_callsite(2, "versions scales with size of history")
817
return self._index.get_versions()
819
def has_version(self, version_id):
820
"""See VersionedFile.has_version."""
821
if 'evil' in debug.debug_flags:
822
trace.mutter_callsite(2, "has_version is a LBYL scenario")
823
return self._index.has_version(version_id)
825
__contains__ = has_version
827
def _merge_annotations(self, content, parents, parent_texts={},
828
delta=None, annotated=None,
829
left_matching_blocks=None):
830
"""Merge annotations for content. This is done by comparing
831
the annotations based on changed to the text.
833
if left_matching_blocks is not None:
834
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
838
for parent_id in parents:
839
merge_content = self._get_content(parent_id, parent_texts)
840
if (parent_id == parents[0] and delta_seq is not None):
843
seq = patiencediff.PatienceSequenceMatcher(
844
None, merge_content.text(), content.text())
845
for i, j, n in seq.get_matching_blocks():
848
# this appears to copy (origin, text) pairs across to the
849
# new content for any line that matches the last-checked
851
content._lines[j:j+n] = merge_content._lines[i:i+n]
853
if delta_seq is None:
854
reference_content = self._get_content(parents[0], parent_texts)
855
new_texts = content.text()
856
old_texts = reference_content.text()
857
delta_seq = patiencediff.PatienceSequenceMatcher(
858
None, old_texts, new_texts)
859
return self._make_line_delta(delta_seq, content)
861
def _make_line_delta(self, delta_seq, new_content):
862
"""Generate a line delta from delta_seq and new_content."""
864
for op in delta_seq.get_opcodes():
867
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
870
def _get_components_positions(self, version_ids):
871
"""Produce a map of position data for the components of versions.
873
This data is intended to be used for retrieving the knit records.
875
A dict of version_id to (method, data_pos, data_size, next) is
877
method is the way referenced data should be applied.
878
data_pos is the position of the data in the knit.
879
data_size is the size of the data in the knit.
880
next is the build-parent of the version, or None for fulltexts.
883
for version_id in version_ids:
886
while cursor is not None and cursor not in component_data:
887
method = self._index.get_method(cursor)
888
if method == 'fulltext':
891
next = self.get_parents_with_ghosts(cursor)[0]
892
index_memo = self._index.get_position(cursor)
893
component_data[cursor] = (method, index_memo, next)
895
return component_data
897
def _get_content(self, version_id, parent_texts={}):
898
"""Returns a content object that makes up the specified
900
cached_version = parent_texts.get(version_id, None)
901
if cached_version is not None:
902
if not self.has_version(version_id):
903
raise RevisionNotPresent(version_id, self.filename)
904
return cached_version
906
text_map, contents_map = self._get_content_maps([version_id])
907
return contents_map[version_id]
909
def _check_versions_present(self, version_ids):
910
"""Check that all specified versions are present."""
911
self._index.check_versions_present(version_ids)
913
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
914
nostore_sha, random_id, check_content):
915
"""See VersionedFile.add_lines_with_ghosts()."""
916
self._check_add(version_id, lines, random_id, check_content)
917
return self._add(version_id, lines, parents, self.delta,
918
parent_texts, None, nostore_sha, random_id)
920
def _add_lines(self, version_id, parents, lines, parent_texts,
921
left_matching_blocks, nostore_sha, random_id, check_content):
922
"""See VersionedFile.add_lines."""
923
self._check_add(version_id, lines, random_id, check_content)
924
self._check_versions_present(parents)
925
return self._add(version_id, lines[:], parents, self.delta,
926
parent_texts, left_matching_blocks, nostore_sha, random_id)
928
def _check_add(self, version_id, lines, random_id, check_content):
929
"""check that version_id and lines are safe to add."""
930
if contains_whitespace(version_id):
931
raise InvalidRevisionId(version_id, self.filename)
932
self.check_not_reserved_id(version_id)
933
# Technically this could be avoided if we are happy to allow duplicate
934
# id insertion when other things than bzr core insert texts, but it
935
# seems useful for folk using the knit api directly to have some safety
936
# blanket that we can disable.
937
if not random_id and self.has_version(version_id):
938
raise RevisionAlreadyPresent(version_id, self.filename)
940
self._check_lines_not_unicode(lines)
941
self._check_lines_are_lines(lines)
943
def _add(self, version_id, lines, parents, delta, parent_texts,
944
left_matching_blocks, nostore_sha, random_id):
945
"""Add a set of lines on top of version specified by parents.
947
If delta is true, compress the text as a line-delta against
950
Any versions not present will be converted into ghosts.
952
# first thing, if the content is something we don't need to store, find
954
line_bytes = ''.join(lines)
955
digest = sha_string(line_bytes)
956
if nostore_sha == digest:
957
raise errors.ExistingContent
960
if parent_texts is None:
962
for parent in parents:
963
if self.has_version(parent):
964
present_parents.append(parent)
966
# can only compress against the left most present parent.
968
(len(present_parents) == 0 or
969
present_parents[0] != parents[0])):
972
text_length = len(line_bytes)
975
if lines[-1][-1] != '\n':
976
# copy the contents of lines.
978
options.append('no-eol')
979
lines[-1] = lines[-1] + '\n'
983
# To speed the extract of texts the delta chain is limited
984
# to a fixed number of deltas. This should minimize both
985
# I/O and the time spend applying deltas.
986
delta = self._check_should_delta(present_parents)
988
assert isinstance(version_id, str)
989
content = self.factory.make(lines, version_id)
990
if delta or (self.factory.annotated and len(present_parents) > 0):
991
# Merge annotations from parent texts if needed.
992
delta_hunks = self._merge_annotations(content, present_parents,
993
parent_texts, delta, self.factory.annotated,
994
left_matching_blocks)
997
options.append('line-delta')
998
store_lines = self.factory.lower_line_delta(delta_hunks)
999
size, bytes = self._data._record_to_data(version_id, digest,
1002
options.append('fulltext')
1003
# isinstance is slower and we have no hierarchy.
1004
if self.factory.__class__ == KnitPlainFactory:
1005
# Use the already joined bytes saving iteration time in
1007
size, bytes = self._data._record_to_data(version_id, digest,
1008
lines, [line_bytes])
1010
# get mixed annotation + content and feed it into the
1012
store_lines = self.factory.lower_fulltext(content)
1013
size, bytes = self._data._record_to_data(version_id, digest,
1016
access_memo = self._data.add_raw_records([size], bytes)[0]
1017
self._index.add_versions(
1018
((version_id, options, access_memo, parents),),
1019
random_id=random_id)
1020
return digest, text_length, content
1022
def check(self, progress_bar=None):
1023
"""See VersionedFile.check()."""
1025
def _clone_text(self, new_version_id, old_version_id, parents):
1026
"""See VersionedFile.clone_text()."""
1027
# FIXME RBC 20060228 make fast by only inserting an index with null
1029
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1031
def get_lines(self, version_id):
1032
"""See VersionedFile.get_lines()."""
1033
return self.get_line_list([version_id])[0]
1035
def _get_record_map(self, version_ids):
1036
"""Produce a dictionary of knit records.
1038
The keys are version_ids, the values are tuples of (method, content,
1040
method is the way the content should be applied.
1041
content is a KnitContent object.
1042
digest is the SHA1 digest of this version id after all steps are done
1043
next is the build-parent of the version, i.e. the leftmost ancestor.
1044
If the method is fulltext, next will be None.
1046
position_map = self._get_components_positions(version_ids)
1047
# c = component_id, m = method, i_m = index_memo, n = next
1048
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1050
for component_id, content, digest in \
1051
self._data.read_records_iter(records):
1052
method, index_memo, next = position_map[component_id]
1053
record_map[component_id] = method, content, digest, next
1057
def get_text(self, version_id):
1058
"""See VersionedFile.get_text"""
1059
return self.get_texts([version_id])[0]
1061
def get_texts(self, version_ids):
1062
return [''.join(l) for l in self.get_line_list(version_ids)]
1064
def get_line_list(self, version_ids):
1065
"""Return the texts of listed versions as a list of strings."""
1066
for version_id in version_ids:
1067
self.check_not_reserved_id(version_id)
1068
text_map, content_map = self._get_content_maps(version_ids)
1069
return [text_map[v] for v in version_ids]
1071
_get_lf_split_line_list = get_line_list
1073
def _get_content_maps(self, version_ids):
1074
"""Produce maps of text and KnitContents
1076
:return: (text_map, content_map) where text_map contains the texts for
1077
the requested versions and content_map contains the KnitContents.
1078
Both dicts take version_ids as their keys.
1080
# FUTURE: This function could be improved for the 'extract many' case
1081
# by tracking each component and only doing the copy when the number of
1082
# children than need to apply delta's to it is > 1 or it is part of the
1084
version_ids = list(version_ids)
1085
multiple_versions = len(version_ids) != 1
1086
record_map = self._get_record_map(version_ids)
1091
for version_id in version_ids:
1094
while cursor is not None:
1095
method, data, digest, next = record_map[cursor]
1096
components.append((cursor, method, data, digest))
1097
if cursor in content_map:
1102
for component_id, method, data, digest in reversed(components):
1103
if component_id in content_map:
1104
content = content_map[component_id]
1106
if method == 'fulltext':
1107
assert content is None
1108
content = self.factory.parse_fulltext(data, version_id)
1109
elif method == 'line-delta':
1110
delta = self.factory.parse_line_delta(data, version_id)
1111
if multiple_versions:
1112
# only doing this when we want multiple versions
1113
# output avoids list copies - which reference and
1114
# dereference many strings.
1115
content = content.copy()
1116
content.apply_delta(delta, version_id)
1117
if multiple_versions:
1118
content_map[component_id] = content
1120
if 'no-eol' in self._index.get_options(version_id):
1121
if multiple_versions:
1122
content = content.copy()
1123
content.strip_last_line_newline()
1124
final_content[version_id] = content
1126
# digest here is the digest from the last applied component.
1127
text = content.text()
1128
actual_sha = sha_strings(text)
1129
if actual_sha != digest:
1130
raise KnitCorrupt(self.filename,
1132
'\n of reconstructed text does not match'
1134
'\n for version %s' %
1135
(actual_sha, digest, version_id))
1136
text_map[version_id] = text
1137
return text_map, final_content
1139
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1141
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1142
if version_ids is None:
1143
version_ids = self.versions()
1145
pb = progress.DummyProgress()
1146
# we don't care about inclusions, the caller cares.
1147
# but we need to setup a list of records to visit.
1148
# we need version_id, position, length
1149
version_id_records = []
1150
requested_versions = set(version_ids)
1151
# filter for available versions
1152
for version_id in requested_versions:
1153
if not self.has_version(version_id):
1154
raise RevisionNotPresent(version_id, self.filename)
1155
# get a in-component-order queue:
1156
for version_id in self.versions():
1157
if version_id in requested_versions:
1158
index_memo = self._index.get_position(version_id)
1159
version_id_records.append((version_id, index_memo))
1161
total = len(version_id_records)
1162
for version_idx, (version_id, data, sha_value) in \
1163
enumerate(self._data.read_records_iter(version_id_records)):
1164
pb.update('Walking content.', version_idx, total)
1165
method = self._index.get_method(version_id)
1167
assert method in ('fulltext', 'line-delta')
1168
if method == 'fulltext':
1169
line_iterator = self.factory.get_fulltext_content(data)
1171
line_iterator = self.factory.get_linedelta_content(data)
1172
# XXX: It might be more efficient to yield (version_id,
1173
# line_iterator) in the future. However for now, this is a simpler
1174
# change to integrate into the rest of the codebase. RBC 20071110
1175
for line in line_iterator:
1176
yield line, version_id
1178
pb.update('Walking content.', total, total)
1180
def iter_parents(self, version_ids):
1181
"""Iterate through the parents for many version ids.
1183
:param version_ids: An iterable yielding version_ids.
1184
:return: An iterator that yields (version_id, parents). Requested
1185
version_ids not present in the versioned file are simply skipped.
1186
The order is undefined, allowing for different optimisations in
1187
the underlying implementation.
1189
return self._index.iter_parents(version_ids)
1191
def num_versions(self):
1192
"""See VersionedFile.num_versions()."""
1193
return self._index.num_versions()
1195
__len__ = num_versions
1197
def annotate_iter(self, version_id):
1198
"""See VersionedFile.annotate_iter."""
1199
return self.factory.annotate_iter(self, version_id)
1201
def get_parents(self, version_id):
1202
"""See VersionedFile.get_parents."""
1205
# 52554 calls in 1264 872 internal down from 3674
1207
return self._index.get_parents(version_id)
1209
raise RevisionNotPresent(version_id, self.filename)
1211
def get_parents_with_ghosts(self, version_id):
1212
"""See VersionedFile.get_parents."""
1214
return self._index.get_parents_with_ghosts(version_id)
1216
raise RevisionNotPresent(version_id, self.filename)
1218
def get_ancestry(self, versions, topo_sorted=True):
1219
"""See VersionedFile.get_ancestry."""
1220
if isinstance(versions, basestring):
1221
versions = [versions]
1224
return self._index.get_ancestry(versions, topo_sorted)
1226
def get_ancestry_with_ghosts(self, versions):
1227
"""See VersionedFile.get_ancestry_with_ghosts."""
1228
if isinstance(versions, basestring):
1229
versions = [versions]
1232
return self._index.get_ancestry_with_ghosts(versions)
1234
def plan_merge(self, ver_a, ver_b):
1235
"""See VersionedFile.plan_merge."""
1236
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1237
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1238
annotated_a = self.annotate(ver_a)
1239
annotated_b = self.annotate(ver_b)
1240
return merge._plan_annotate_merge(annotated_a, annotated_b,
1241
ancestors_a, ancestors_b)
1244
class _KnitComponentFile(object):
1245
"""One of the files used to implement a knit database"""
1247
def __init__(self, transport, filename, mode, file_mode=None,
1248
create_parent_dir=False, dir_mode=None):
1249
self._transport = transport
1250
self._filename = filename
1252
self._file_mode = file_mode
1253
self._dir_mode = dir_mode
1254
self._create_parent_dir = create_parent_dir
1255
self._need_to_create = False
1257
def _full_path(self):
1258
"""Return the full path to this file."""
1259
return self._transport.base + self._filename
1261
def check_header(self, fp):
1262
line = fp.readline()
1264
# An empty file can actually be treated as though the file doesn't
1266
raise errors.NoSuchFile(self._full_path())
1267
if line != self.HEADER:
1268
raise KnitHeaderError(badline=line,
1269
filename=self._transport.abspath(self._filename))
1272
return '%s(%s)' % (self.__class__.__name__, self._filename)
1275
class _KnitIndex(_KnitComponentFile):
1276
"""Manages knit index file.
1278
The index is already kept in memory and read on startup, to enable
1279
fast lookups of revision information. The cursor of the index
1280
file is always pointing to the end, making it easy to append
1283
_cache is a cache for fast mapping from version id to a Index
1286
_history is a cache for fast mapping from indexes to version ids.
1288
The index data format is dictionary compressed when it comes to
1289
parent references; a index entry may only have parents that with a
1290
lover index number. As a result, the index is topological sorted.
1292
Duplicate entries may be written to the index for a single version id
1293
if this is done then the latter one completely replaces the former:
1294
this allows updates to correct version and parent information.
1295
Note that the two entries may share the delta, and that successive
1296
annotations and references MUST point to the first entry.
1298
The index file on disc contains a header, followed by one line per knit
1299
record. The same revision can be present in an index file more than once.
1300
The first occurrence gets assigned a sequence number starting from 0.
1302
The format of a single line is
1303
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1304
REVISION_ID is a utf8-encoded revision id
1305
FLAGS is a comma separated list of flags about the record. Values include
1306
no-eol, line-delta, fulltext.
1307
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1308
that the the compressed data starts at.
1309
LENGTH is the ascii representation of the length of the data file.
1310
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1312
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1313
revision id already in the knit that is a parent of REVISION_ID.
1314
The ' :' marker is the end of record marker.
1317
when a write is interrupted to the index file, it will result in a line
1318
that does not end in ' :'. If the ' :' is not present at the end of a line,
1319
or at the end of the file, then the record that is missing it will be
1320
ignored by the parser.
1322
When writing new records to the index file, the data is preceded by '\n'
1323
to ensure that records always start on new lines even if the last write was
1324
interrupted. As a result its normal for the last line in the index to be
1325
missing a trailing newline. One can be added with no harmful effects.
1328
HEADER = "# bzr knit index 8\n"
1330
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1331
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1333
def _cache_version(self, version_id, options, pos, size, parents):
1334
"""Cache a version record in the history array and index cache.
1336
This is inlined into _load_data for performance. KEEP IN SYNC.
1337
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1340
# only want the _history index to reference the 1st index entry
1342
if version_id not in self._cache:
1343
index = len(self._history)
1344
self._history.append(version_id)
1346
index = self._cache[version_id][5]
1347
self._cache[version_id] = (version_id,
1354
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1355
create_parent_dir=False, delay_create=False, dir_mode=None):
1356
_KnitComponentFile.__init__(self, transport, filename, mode,
1357
file_mode=file_mode,
1358
create_parent_dir=create_parent_dir,
1361
# position in _history is the 'official' index for a revision
1362
# but the values may have come from a newer entry.
1363
# so - wc -l of a knit index is != the number of unique names
1367
fp = self._transport.get(self._filename)
1369
# _load_data may raise NoSuchFile if the target knit is
1371
_load_data(self, fp)
1375
if mode != 'w' or not create:
1378
self._need_to_create = True
1380
self._transport.put_bytes_non_atomic(
1381
self._filename, self.HEADER, mode=self._file_mode)
1383
def get_graph(self):
1384
"""Return a list of the node:parents lists from this knit index."""
1385
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1387
def get_ancestry(self, versions, topo_sorted=True):
1388
"""See VersionedFile.get_ancestry."""
1389
# get a graph of all the mentioned versions:
1391
pending = set(versions)
1394
version = pending.pop()
1397
parents = [p for p in cache[version][4] if p in cache]
1399
raise RevisionNotPresent(version, self._filename)
1400
# if not completed and not a ghost
1401
pending.update([p for p in parents if p not in graph])
1402
graph[version] = parents
1405
return topo_sort(graph.items())
1407
def get_ancestry_with_ghosts(self, versions):
1408
"""See VersionedFile.get_ancestry_with_ghosts."""
1409
# get a graph of all the mentioned versions:
1410
self.check_versions_present(versions)
1413
pending = set(versions)
1415
version = pending.pop()
1417
parents = cache[version][4]
1423
pending.update([p for p in parents if p not in graph])
1424
graph[version] = parents
1425
return topo_sort(graph.items())
1427
def iter_parents(self, version_ids):
1428
"""Iterate through the parents for many version ids.
1430
:param version_ids: An iterable yielding version_ids.
1431
:return: An iterator that yields (version_id, parents). Requested
1432
version_ids not present in the versioned file are simply skipped.
1433
The order is undefined, allowing for different optimisations in
1434
the underlying implementation.
1436
for version_id in version_ids:
1438
yield version_id, tuple(self.get_parents(version_id))
1442
def num_versions(self):
1443
return len(self._history)
1445
__len__ = num_versions
1447
def get_versions(self):
1448
"""Get all the versions in the file. not topologically sorted."""
1449
return self._history
1451
def _version_list_to_index(self, versions):
1454
for version in versions:
1455
if version in cache:
1456
# -- inlined lookup() --
1457
result_list.append(str(cache[version][5]))
1458
# -- end lookup () --
1460
result_list.append('.' + version)
1461
return ' '.join(result_list)
1463
def add_version(self, version_id, options, index_memo, parents):
1464
"""Add a version record to the index."""
1465
self.add_versions(((version_id, options, index_memo, parents),))
1467
def add_versions(self, versions, random_id=False):
1468
"""Add multiple versions to the index.
1470
:param versions: a list of tuples:
1471
(version_id, options, pos, size, parents).
1472
:param random_id: If True the ids being added were randomly generated
1473
and no check for existence will be performed.
1476
orig_history = self._history[:]
1477
orig_cache = self._cache.copy()
1480
for version_id, options, (index, pos, size), parents in versions:
1481
line = "\n%s %s %s %s %s :" % (version_id,
1485
self._version_list_to_index(parents))
1486
assert isinstance(line, str), \
1487
'content must be utf-8 encoded: %r' % (line,)
1489
self._cache_version(version_id, options, pos, size, parents)
1490
if not self._need_to_create:
1491
self._transport.append_bytes(self._filename, ''.join(lines))
1494
sio.write(self.HEADER)
1495
sio.writelines(lines)
1497
self._transport.put_file_non_atomic(self._filename, sio,
1498
create_parent_dir=self._create_parent_dir,
1499
mode=self._file_mode,
1500
dir_mode=self._dir_mode)
1501
self._need_to_create = False
1503
# If any problems happen, restore the original values and re-raise
1504
self._history = orig_history
1505
self._cache = orig_cache
1508
def has_version(self, version_id):
1509
"""True if the version is in the index."""
1510
return version_id in self._cache
1512
def get_position(self, version_id):
1513
"""Return details needed to access the version.
1515
.kndx indices do not support split-out data, so return None for the
1518
:return: a tuple (None, data position, size) to hand to the access
1519
logic to get the record.
1521
entry = self._cache[version_id]
1522
return None, entry[2], entry[3]
1524
def get_method(self, version_id):
1525
"""Return compression method of specified version."""
1527
options = self._cache[version_id][1]
1529
raise RevisionNotPresent(version_id, self._filename)
1530
if 'fulltext' in options:
1533
if 'line-delta' not in options:
1534
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1537
def get_options(self, version_id):
1538
"""Return a list representing options.
1542
return self._cache[version_id][1]
1544
def get_parents(self, version_id):
1545
"""Return parents of specified version ignoring ghosts."""
1546
return [parent for parent in self._cache[version_id][4]
1547
if parent in self._cache]
1549
def get_parents_with_ghosts(self, version_id):
1550
"""Return parents of specified version with ghosts."""
1551
return self._cache[version_id][4]
1553
def check_versions_present(self, version_ids):
1554
"""Check that all specified versions are present."""
1556
for version_id in version_ids:
1557
if version_id not in cache:
1558
raise RevisionNotPresent(version_id, self._filename)
1561
class KnitGraphIndex(object):
1562
"""A knit index that builds on GraphIndex."""
1564
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1565
"""Construct a KnitGraphIndex on a graph_index.
1567
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1568
:param deltas: Allow delta-compressed records.
1569
:param add_callback: If not None, allow additions to the index and call
1570
this callback with a list of added GraphIndex nodes:
1571
[(node, value, node_refs), ...]
1572
:param parents: If True, record knits parents, if not do not record
1575
self._graph_index = graph_index
1576
self._deltas = deltas
1577
self._add_callback = add_callback
1578
self._parents = parents
1579
if deltas and not parents:
1580
raise KnitCorrupt(self, "Cannot do delta compression without "
1583
def _get_entries(self, keys, check_present=False):
1584
"""Get the entries for keys.
1586
:param keys: An iterable of index keys, - 1-tuples.
1591
for node in self._graph_index.iter_entries(keys):
1593
found_keys.add(node[1])
1595
# adapt parentless index to the rest of the code.
1596
for node in self._graph_index.iter_entries(keys):
1597
yield node[0], node[1], node[2], ()
1598
found_keys.add(node[1])
1600
missing_keys = keys.difference(found_keys)
1602
raise RevisionNotPresent(missing_keys.pop(), self)
1604
def _present_keys(self, version_ids):
1606
node[1] for node in self._get_entries(version_ids)])
1608
def _parentless_ancestry(self, versions):
1609
"""Honour the get_ancestry API for parentless knit indices."""
1610
wanted_keys = self._version_ids_to_keys(versions)
1611
present_keys = self._present_keys(wanted_keys)
1612
missing = set(wanted_keys).difference(present_keys)
1614
raise RevisionNotPresent(missing.pop(), self)
1615
return list(self._keys_to_version_ids(present_keys))
1617
def get_ancestry(self, versions, topo_sorted=True):
1618
"""See VersionedFile.get_ancestry."""
1619
if not self._parents:
1620
return self._parentless_ancestry(versions)
1621
# XXX: This will do len(history) index calls - perhaps
1622
# it should be altered to be a index core feature?
1623
# get a graph of all the mentioned versions:
1626
versions = self._version_ids_to_keys(versions)
1627
pending = set(versions)
1629
# get all pending nodes
1630
this_iteration = pending
1631
new_nodes = self._get_entries(this_iteration)
1634
for (index, key, value, node_refs) in new_nodes:
1635
# dont ask for ghosties - otherwise
1636
# we we can end up looping with pending
1637
# being entirely ghosted.
1638
graph[key] = [parent for parent in node_refs[0]
1639
if parent not in ghosts]
1641
for parent in graph[key]:
1642
# dont examine known nodes again
1647
ghosts.update(this_iteration.difference(found))
1648
if versions.difference(graph):
1649
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1651
result_keys = topo_sort(graph.items())
1653
result_keys = graph.iterkeys()
1654
return [key[0] for key in result_keys]
1656
def get_ancestry_with_ghosts(self, versions):
1657
"""See VersionedFile.get_ancestry."""
1658
if not self._parents:
1659
return self._parentless_ancestry(versions)
1660
# XXX: This will do len(history) index calls - perhaps
1661
# it should be altered to be a index core feature?
1662
# get a graph of all the mentioned versions:
1664
versions = self._version_ids_to_keys(versions)
1665
pending = set(versions)
1667
# get all pending nodes
1668
this_iteration = pending
1669
new_nodes = self._get_entries(this_iteration)
1671
for (index, key, value, node_refs) in new_nodes:
1672
graph[key] = node_refs[0]
1674
for parent in graph[key]:
1675
# dont examine known nodes again
1679
missing_versions = this_iteration.difference(graph)
1680
missing_needed = versions.intersection(missing_versions)
1682
raise RevisionNotPresent(missing_needed.pop(), self)
1683
for missing_version in missing_versions:
1684
# add a key, no parents
1685
graph[missing_version] = []
1686
pending.discard(missing_version) # don't look for it
1687
result_keys = topo_sort(graph.items())
1688
return [key[0] for key in result_keys]
1690
def get_graph(self):
1691
"""Return a list of the node:parents lists from this knit index."""
1692
if not self._parents:
1693
return [(key, ()) for key in self.get_versions()]
1695
for index, key, value, refs in self._graph_index.iter_all_entries():
1696
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1699
def iter_parents(self, version_ids):
1700
"""Iterate through the parents for many version ids.
1702
:param version_ids: An iterable yielding version_ids.
1703
:return: An iterator that yields (version_id, parents). Requested
1704
version_ids not present in the versioned file are simply skipped.
1705
The order is undefined, allowing for different optimisations in
1706
the underlying implementation.
1709
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1711
present_parents = set()
1712
for node in all_nodes:
1713
all_parents.update(node[3][0])
1714
# any node we are querying must be present
1715
present_parents.add(node[1])
1716
unknown_parents = all_parents.difference(present_parents)
1717
present_parents.update(self._present_keys(unknown_parents))
1718
for node in all_nodes:
1720
for parent in node[3][0]:
1721
if parent in present_parents:
1722
parents.append(parent[0])
1723
yield node[1][0], tuple(parents)
1725
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1726
yield node[1][0], ()
1728
def num_versions(self):
1729
return len(list(self._graph_index.iter_all_entries()))
1731
__len__ = num_versions
1733
def get_versions(self):
1734
"""Get all the versions in the file. not topologically sorted."""
1735
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1737
def has_version(self, version_id):
1738
"""True if the version is in the index."""
1739
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1741
def _keys_to_version_ids(self, keys):
1742
return tuple(key[0] for key in keys)
1744
def get_position(self, version_id):
1745
"""Return details needed to access the version.
1747
:return: a tuple (index, data position, size) to hand to the access
1748
logic to get the record.
1750
node = self._get_node(version_id)
1751
bits = node[2][1:].split(' ')
1752
return node[0], int(bits[0]), int(bits[1])
1754
def get_method(self, version_id):
1755
"""Return compression method of specified version."""
1756
if not self._deltas:
1758
return self._parent_compression(self._get_node(version_id)[3][1])
1760
def _parent_compression(self, reference_list):
1761
# use the second reference list to decide if this is delta'd or not.
1762
if len(reference_list):
1767
def _get_node(self, version_id):
1769
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1771
raise RevisionNotPresent(version_id, self)
1773
def get_options(self, version_id):
1774
"""Return a list representing options.
1778
node = self._get_node(version_id)
1779
if not self._deltas:
1780
options = ['fulltext']
1782
options = [self._parent_compression(node[3][1])]
1783
if node[2][0] == 'N':
1784
options.append('no-eol')
1787
def get_parents(self, version_id):
1788
"""Return parents of specified version ignoring ghosts."""
1789
parents = list(self.iter_parents([version_id]))
1792
raise errors.RevisionNotPresent(version_id, self)
1793
return parents[0][1]
1795
def get_parents_with_ghosts(self, version_id):
1796
"""Return parents of specified version with ghosts."""
1797
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1798
check_present=True))
1799
if not self._parents:
1801
return self._keys_to_version_ids(nodes[0][3][0])
1803
def check_versions_present(self, version_ids):
1804
"""Check that all specified versions are present."""
1805
keys = self._version_ids_to_keys(version_ids)
1806
present = self._present_keys(keys)
1807
missing = keys.difference(present)
1809
raise RevisionNotPresent(missing.pop(), self)
1811
def add_version(self, version_id, options, access_memo, parents):
1812
"""Add a version record to the index."""
1813
return self.add_versions(((version_id, options, access_memo, parents),))
1815
def add_versions(self, versions, random_id=False):
1816
"""Add multiple versions to the index.
1818
This function does not insert data into the Immutable GraphIndex
1819
backing the KnitGraphIndex, instead it prepares data for insertion by
1820
the caller and checks that it is safe to insert then calls
1821
self._add_callback with the prepared GraphIndex nodes.
1823
:param versions: a list of tuples:
1824
(version_id, options, pos, size, parents).
1825
:param random_id: If True the ids being added were randomly generated
1826
and no check for existence will be performed.
1828
if not self._add_callback:
1829
raise errors.ReadOnlyError(self)
1830
# we hope there are no repositories with inconsistent parentage
1835
for (version_id, options, access_memo, parents) in versions:
1836
index, pos, size = access_memo
1837
key = (version_id, )
1838
parents = tuple((parent, ) for parent in parents)
1839
if 'no-eol' in options:
1843
value += "%d %d" % (pos, size)
1844
if not self._deltas:
1845
if 'line-delta' in options:
1846
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1849
if 'line-delta' in options:
1850
node_refs = (parents, (parents[0],))
1852
node_refs = (parents, ())
1854
node_refs = (parents, )
1857
raise KnitCorrupt(self, "attempt to add node with parents "
1858
"in parentless index.")
1860
keys[key] = (value, node_refs)
1862
present_nodes = self._get_entries(keys)
1863
for (index, key, value, node_refs) in present_nodes:
1864
if (value, node_refs) != keys[key]:
1865
raise KnitCorrupt(self, "inconsistent details in add_versions"
1866
": %s %s" % ((value, node_refs), keys[key]))
1870
for key, (value, node_refs) in keys.iteritems():
1871
result.append((key, value, node_refs))
1873
for key, (value, node_refs) in keys.iteritems():
1874
result.append((key, value))
1875
self._add_callback(result)
1877
def _version_ids_to_keys(self, version_ids):
1878
return set((version_id, ) for version_id in version_ids)
1881
class _KnitAccess(object):
1882
"""Access to knit records in a .knit file."""
1884
def __init__(self, transport, filename, _file_mode, _dir_mode,
1885
_need_to_create, _create_parent_dir):
1886
"""Create a _KnitAccess for accessing and inserting data.
1888
:param transport: The transport the .knit is located on.
1889
:param filename: The filename of the .knit.
1891
self._transport = transport
1892
self._filename = filename
1893
self._file_mode = _file_mode
1894
self._dir_mode = _dir_mode
1895
self._need_to_create = _need_to_create
1896
self._create_parent_dir = _create_parent_dir
1898
def add_raw_records(self, sizes, raw_data):
1899
"""Add raw knit bytes to a storage area.
1901
The data is spooled to whereever the access method is storing data.
1903
:param sizes: An iterable containing the size of each raw data segment.
1904
:param raw_data: A bytestring containing the data.
1905
:return: A list of memos to retrieve the record later. Each memo is a
1906
tuple - (index, pos, length), where the index field is always None
1907
for the .knit access method.
1909
assert type(raw_data) == str, \
1910
'data must be plain bytes was %s' % type(raw_data)
1911
if not self._need_to_create:
1912
base = self._transport.append_bytes(self._filename, raw_data)
1914
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1915
create_parent_dir=self._create_parent_dir,
1916
mode=self._file_mode,
1917
dir_mode=self._dir_mode)
1918
self._need_to_create = False
1922
result.append((None, base, size))
1927
"""IFF this data access has its own storage area, initialise it.
1931
self._transport.put_bytes_non_atomic(self._filename, '',
1932
mode=self._file_mode)
1934
def open_file(self):
1935
"""IFF this data access can be represented as a single file, open it.
1937
For knits that are not mapped to a single file on disk this will
1940
:return: None or a file handle.
1943
return self._transport.get(self._filename)
1948
def get_raw_records(self, memos_for_retrieval):
1949
"""Get the raw bytes for a records.
1951
:param memos_for_retrieval: An iterable containing the (index, pos,
1952
length) memo for retrieving the bytes. The .knit method ignores
1953
the index as there is always only a single file.
1954
:return: An iterator over the bytes of the records.
1956
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1957
for pos, data in self._transport.readv(self._filename, read_vector):
1961
class _PackAccess(object):
1962
"""Access to knit records via a collection of packs."""
1964
def __init__(self, index_to_packs, writer=None):
1965
"""Create a _PackAccess object.
1967
:param index_to_packs: A dict mapping index objects to the transport
1968
and file names for obtaining data.
1969
:param writer: A tuple (pack.ContainerWriter, write_index) which
1970
contains the pack to write, and the index that reads from it will
1974
self.container_writer = writer[0]
1975
self.write_index = writer[1]
1977
self.container_writer = None
1978
self.write_index = None
1979
self.indices = index_to_packs
1981
def add_raw_records(self, sizes, raw_data):
1982
"""Add raw knit bytes to a storage area.
1984
The data is spooled to the container writer in one bytes-record per
1987
:param sizes: An iterable containing the size of each raw data segment.
1988
:param raw_data: A bytestring containing the data.
1989
:return: A list of memos to retrieve the record later. Each memo is a
1990
tuple - (index, pos, length), where the index field is the
1991
write_index object supplied to the PackAccess object.
1993
assert type(raw_data) == str, \
1994
'data must be plain bytes was %s' % type(raw_data)
1998
p_offset, p_length = self.container_writer.add_bytes_record(
1999
raw_data[offset:offset+size], [])
2001
result.append((self.write_index, p_offset, p_length))
2005
"""Pack based knits do not get individually created."""
2007
def get_raw_records(self, memos_for_retrieval):
2008
"""Get the raw bytes for a records.
2010
:param memos_for_retrieval: An iterable containing the (index, pos,
2011
length) memo for retrieving the bytes. The Pack access method
2012
looks up the pack to use for a given record in its index_to_pack
2014
:return: An iterator over the bytes of the records.
2016
# first pass, group into same-index requests
2018
current_index = None
2019
for (index, offset, length) in memos_for_retrieval:
2020
if current_index == index:
2021
current_list.append((offset, length))
2023
if current_index is not None:
2024
request_lists.append((current_index, current_list))
2025
current_index = index
2026
current_list = [(offset, length)]
2027
# handle the last entry
2028
if current_index is not None:
2029
request_lists.append((current_index, current_list))
2030
for index, offsets in request_lists:
2031
transport, path = self.indices[index]
2032
reader = pack.make_readv_reader(transport, path, offsets)
2033
for names, read_func in reader.iter_records():
2034
yield read_func(None)
2036
def open_file(self):
2037
"""Pack based knits have no single file."""
2040
def set_writer(self, writer, index, (transport, packname)):
2041
"""Set a writer to use for adding data."""
2042
if index is not None:
2043
self.indices[index] = (transport, packname)
2044
self.container_writer = writer
2045
self.write_index = index
2048
class _StreamAccess(object):
2049
"""A Knit Access object that provides data from a datastream.
2051
It also provides a fallback to present as unannotated data, annotated data
2052
from a *backing* access object.
2054
This is triggered by a index_memo which is pointing to a different index
2055
than this was constructed with, and is used to allow extracting full
2056
unannotated texts for insertion into annotated knits.
2059
def __init__(self, reader_callable, stream_index, backing_knit,
2061
"""Create a _StreamAccess object.
2063
:param reader_callable: The reader_callable from the datastream.
2064
This is called to buffer all the data immediately, for
2066
:param stream_index: The index the data stream this provides access to
2067
which will be present in native index_memo's.
2068
:param backing_knit: The knit object that will provide access to
2069
annotated texts which are not available in the stream, so as to
2070
create unannotated texts.
2071
:param orig_factory: The original content factory used to generate the
2072
stream. This is used for checking whether the thunk code for
2073
supporting _copy_texts will generate the correct form of data.
2075
self.data = reader_callable(None)
2076
self.stream_index = stream_index
2077
self.backing_knit = backing_knit
2078
self.orig_factory = orig_factory
2080
def get_raw_records(self, memos_for_retrieval):
2081
"""Get the raw bytes for a records.
2083
:param memos_for_retrieval: An iterable containing the (thunk_flag,
2084
index, start, end) memo for retrieving the bytes.
2085
:return: An iterator over the bytes of the records.
2087
# use a generator for memory friendliness
2088
for thunk_flag, version_id, start, end in memos_for_retrieval:
2089
if version_id is self.stream_index:
2090
yield self.data[start:end]
2092
# we have been asked to thunk. This thunking only occurs when
2093
# we are obtaining plain texts from an annotated backing knit
2094
# so that _copy_texts will work.
2095
# We could improve performance here by scanning for where we need
2096
# to do this and using get_line_list, then interleaving the output
2097
# as desired. However, for now, this is sufficient.
2098
if self.orig_factory.__class__ != KnitPlainFactory:
2099
raise errors.KnitCorrupt(
2100
self, 'Bad thunk request %r' % version_id)
2101
lines = self.backing_knit.get_lines(version_id)
2102
line_bytes = ''.join(lines)
2103
digest = sha_string(line_bytes)
2105
if lines[-1][-1] != '\n':
2106
lines[-1] = lines[-1] + '\n'
2108
orig_options = list(self.backing_knit._index.get_options(version_id))
2109
if 'fulltext' not in orig_options:
2110
if 'line-delta' not in orig_options:
2111
raise errors.KnitCorrupt(self,
2112
'Unknown compression method %r' % orig_options)
2113
orig_options.remove('line-delta')
2114
orig_options.append('fulltext')
2115
# We want plain data, because we expect to thunk only to allow text
2117
size, bytes = self.backing_knit._data._record_to_data(version_id,
2118
digest, lines, line_bytes)
2122
class _StreamIndex(object):
2123
"""A Knit Index object that uses the data map from a datastream."""
2125
def __init__(self, data_list):
2126
"""Create a _StreamIndex object.
2128
:param data_list: The data_list from the datastream.
2130
self.data_list = data_list
2131
self._by_version = {}
2133
for key, options, length, parents in data_list:
2134
self._by_version[key] = options, (pos, pos + length), parents
2137
def get_ancestry(self, versions, topo_sorted):
2138
"""Get an ancestry list for versions."""
2140
# Not needed for basic joins
2141
raise NotImplementedError(self.get_ancestry)
2142
# get a graph of all the mentioned versions:
2143
# Little ugly - basically copied from KnitIndex, but don't want to
2144
# accidentally incorporate too much of that index's code.
2146
pending = set(versions)
2147
cache = self._by_version
2149
version = pending.pop()
2152
parents = [p for p in cache[version][2] if p in cache]
2154
raise RevisionNotPresent(version, self)
2155
# if not completed and not a ghost
2156
pending.update([p for p in parents if p not in ancestry])
2157
ancestry.add(version)
2158
return list(ancestry)
2160
def get_method(self, version_id):
2161
"""Return compression method of specified version."""
2163
options = self._by_version[version_id][0]
2165
# Strictly speaking this should check in the backing knit, but
2166
# until we have a test to discriminate, this will do.
2168
if 'fulltext' in options:
2170
elif 'line-delta' in options:
2173
raise errors.KnitIndexUnknownMethod(self, options)
2175
def get_options(self, version_id):
2176
"""Return a list representing options.
2180
return self._by_version[version_id][0]
2182
def get_parents_with_ghosts(self, version_id):
2183
"""Return parents of specified version with ghosts."""
2184
return self._by_version[version_id][2]
2186
def get_position(self, version_id):
2187
"""Return details needed to access the version.
2189
_StreamAccess has the data as a big array, so we return slice
2190
coordinates into that (as index_memo's are opaque outside the
2191
index and matching access class).
2193
:return: a tuple (thunk_flag, index, start, end). If thunk_flag is
2194
False, index will be self, otherwise it will be a version id.
2197
start, end = self._by_version[version_id][1]
2198
return False, self, start, end
2200
# Signal to the access object to handle this from the backing knit.
2201
return (True, version_id, None, None)
2203
def get_versions(self):
2204
"""Get all the versions in the stream."""
2205
return self._by_version.keys()
2207
def iter_parents(self, version_ids):
2208
"""Iterate through the parents for many version ids.
2210
:param version_ids: An iterable yielding version_ids.
2211
:return: An iterator that yields (version_id, parents). Requested
2212
version_ids not present in the versioned file are simply skipped.
2213
The order is undefined, allowing for different optimisations in
2214
the underlying implementation.
2217
for version in version_ids:
2219
result.append((version, self._by_version[version][2]))
2225
class _KnitData(object):
2226
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2228
The KnitData class provides the logic for parsing and using knit records,
2229
making use of an access method for the low level read and write operations.
2232
def __init__(self, access):
2233
"""Create a KnitData object.
2235
:param access: The access method to use. Access methods such as
2236
_KnitAccess manage the insertion of raw records and the subsequent
2237
retrieval of the same.
2239
self._access = access
2240
self._checked = False
2241
# TODO: jam 20060713 conceptually, this could spill to disk
2242
# if the cached size gets larger than a certain amount
2243
# but it complicates the model a bit, so for now just use
2244
# a simple dictionary
2246
self._do_cache = False
2248
def enable_cache(self):
2249
"""Enable caching of reads."""
2250
self._do_cache = True
2252
def clear_cache(self):
2253
"""Clear the record cache."""
2254
self._do_cache = False
2257
def _open_file(self):
2258
return self._access.open_file()
2260
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2261
"""Convert version_id, digest, lines into a raw data block.
2263
:param dense_lines: The bytes of lines but in a denser form. For
2264
instance, if lines is a list of 1000 bytestrings each ending in \n,
2265
dense_lines may be a list with one line in it, containing all the
2266
1000's lines and their \n's. Using dense_lines if it is already
2267
known is a win because the string join to create bytes in this
2268
function spends less time resizing the final string.
2269
:return: (len, a StringIO instance with the raw data ready to read.)
2271
# Note: using a string copy here increases memory pressure with e.g.
2272
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2273
# when doing the initial commit of a mozilla tree. RBC 20070921
2274
bytes = ''.join(chain(
2275
["version %s %d %s\n" % (version_id,
2278
dense_lines or lines,
2279
["end %s\n" % version_id]))
2280
assert bytes.__class__ == str
2281
compressed_bytes = bytes_to_gzip(bytes)
2282
return len(compressed_bytes), compressed_bytes
2284
def add_raw_records(self, sizes, raw_data):
2285
"""Append a prepared record to the data file.
2287
:param sizes: An iterable containing the size of each raw data segment.
2288
:param raw_data: A bytestring containing the data.
2289
:return: a list of index data for the way the data was stored.
2290
See the access method add_raw_records documentation for more
2293
return self._access.add_raw_records(sizes, raw_data)
2295
def _parse_record_header(self, version_id, raw_data):
2296
"""Parse a record header for consistency.
2298
:return: the header and the decompressor stream.
2299
as (stream, header_record)
2301
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2303
rec = self._check_header(version_id, df.readline())
2304
except Exception, e:
2305
raise KnitCorrupt(self._access,
2306
"While reading {%s} got %s(%s)"
2307
% (version_id, e.__class__.__name__, str(e)))
2310
def _check_header(self, version_id, line):
2313
raise KnitCorrupt(self._access,
2314
'unexpected number of elements in record header')
2315
if rec[1] != version_id:
2316
raise KnitCorrupt(self._access,
2317
'unexpected version, wanted %r, got %r'
2318
% (version_id, rec[1]))
2321
def _parse_record(self, version_id, data):
2323
# 4168 calls in 2880 217 internal
2324
# 4168 calls to _parse_record_header in 2121
2325
# 4168 calls to readlines in 330
2326
df = GzipFile(mode='rb', fileobj=StringIO(data))
2329
record_contents = df.readlines()
2330
except Exception, e:
2331
raise KnitCorrupt(self._access,
2332
"While reading {%s} got %s(%s)"
2333
% (version_id, e.__class__.__name__, str(e)))
2334
header = record_contents.pop(0)
2335
rec = self._check_header(version_id, header)
2337
last_line = record_contents.pop()
2338
if len(record_contents) != int(rec[2]):
2339
raise KnitCorrupt(self._access,
2340
'incorrect number of lines %s != %s'
2342
% (len(record_contents), int(rec[2]),
2344
if last_line != 'end %s\n' % rec[1]:
2345
raise KnitCorrupt(self._access,
2346
'unexpected version end line %r, wanted %r'
2347
% (last_line, version_id))
2349
return record_contents, rec[3]
2351
def read_records_iter_raw(self, records):
2352
"""Read text records from data file and yield raw data.
2354
This unpacks enough of the text record to validate the id is
2355
as expected but thats all.
2357
# setup an iterator of the external records:
2358
# uses readv so nice and fast we hope.
2360
# grab the disk data needed.
2362
# Don't check _cache if it is empty
2363
needed_offsets = [index_memo for version_id, index_memo
2365
if version_id not in self._cache]
2367
needed_offsets = [index_memo for version_id, index_memo
2370
raw_records = self._access.get_raw_records(needed_offsets)
2372
for version_id, index_memo in records:
2373
if version_id in self._cache:
2374
# This data has already been validated
2375
data = self._cache[version_id]
2377
data = raw_records.next()
2379
self._cache[version_id] = data
2381
# validate the header
2382
df, rec = self._parse_record_header(version_id, data)
2384
yield version_id, data
2386
def read_records_iter(self, records):
2387
"""Read text records from data file and yield result.
2389
The result will be returned in whatever is the fastest to read.
2390
Not by the order requested. Also, multiple requests for the same
2391
record will only yield 1 response.
2392
:param records: A list of (version_id, pos, len) entries
2393
:return: Yields (version_id, contents, digest) in the order
2394
read, not the order requested
2400
# Skip records we have alread seen
2401
yielded_records = set()
2402
needed_records = set()
2403
for record in records:
2404
if record[0] in self._cache:
2405
if record[0] in yielded_records:
2407
yielded_records.add(record[0])
2408
data = self._cache[record[0]]
2409
content, digest = self._parse_record(record[0], data)
2410
yield (record[0], content, digest)
2412
needed_records.add(record)
2413
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2415
needed_records = sorted(set(records), key=operator.itemgetter(1))
2417
if not needed_records:
2420
# The transport optimizes the fetching as well
2421
# (ie, reads continuous ranges.)
2422
raw_data = self._access.get_raw_records(
2423
[index_memo for version_id, index_memo in needed_records])
2425
for (version_id, index_memo), data in \
2426
izip(iter(needed_records), raw_data):
2427
content, digest = self._parse_record(version_id, data)
2429
self._cache[version_id] = data
2430
yield version_id, content, digest
2432
def read_records(self, records):
2433
"""Read records into a dictionary."""
2435
for record_id, content, digest in \
2436
self.read_records_iter(records):
2437
components[record_id] = (content, digest)
2441
class InterKnit(InterVersionedFile):
2442
"""Optimised code paths for knit to knit operations."""
2444
_matching_file_from_factory = KnitVersionedFile
2445
_matching_file_to_factory = KnitVersionedFile
2448
def is_compatible(source, target):
2449
"""Be compatible with knits. """
2451
return (isinstance(source, KnitVersionedFile) and
2452
isinstance(target, KnitVersionedFile))
2453
except AttributeError:
2456
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2457
"""Copy texts to the target by extracting and adding them one by one.
2459
see join() for the parameter definitions.
2461
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2462
graph = self.source.get_graph(version_ids)
2463
order = topo_sort(graph.items())
2465
def size_of_content(content):
2466
return sum(len(line) for line in content.text())
2467
# Cache at most 10MB of parent texts
2468
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2469
compute_size=size_of_content)
2470
# TODO: jam 20071116 It would be nice to have a streaming interface to
2471
# get multiple texts from a source. The source could be smarter
2472
# about how it handled intermediate stages.
2473
# get_line_list() or make_mpdiffs() seem like a possibility, but
2474
# at the moment they extract all full texts into memory, which
2475
# causes us to store more than our 3x fulltext goal.
2476
# Repository.iter_files_bytes() may be another possibility
2477
to_process = [version for version in order
2478
if version not in self.target]
2479
total = len(to_process)
2480
pb = ui.ui_factory.nested_progress_bar()
2482
for index, version in enumerate(to_process):
2483
pb.update('Converting versioned data', index, total)
2484
sha1, num_bytes, parent_text = self.target.add_lines(version,
2485
self.source.get_parents_with_ghosts(version),
2486
self.source.get_lines(version),
2487
parent_texts=parent_cache)
2488
parent_cache[version] = parent_text
2493
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2494
"""See InterVersionedFile.join."""
2495
assert isinstance(self.source, KnitVersionedFile)
2496
assert isinstance(self.target, KnitVersionedFile)
2498
# If the source and target are mismatched w.r.t. annotations vs
2499
# plain, the data needs to be converted accordingly
2500
if self.source.factory.annotated == self.target.factory.annotated:
2502
elif self.source.factory.annotated:
2503
converter = self._anno_to_plain_converter
2505
# We're converting from a plain to an annotated knit. Copy them
2506
# across by full texts.
2507
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2509
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2513
pb = ui.ui_factory.nested_progress_bar()
2515
version_ids = list(version_ids)
2516
if None in version_ids:
2517
version_ids.remove(None)
2519
self.source_ancestry = set(self.source.get_ancestry(version_ids,
2521
this_versions = set(self.target._index.get_versions())
2522
# XXX: For efficiency we should not look at the whole index,
2523
# we only need to consider the referenced revisions - they
2524
# must all be present, or the method must be full-text.
2525
# TODO, RBC 20070919
2526
needed_versions = self.source_ancestry - this_versions
2528
if not needed_versions:
2530
full_list = topo_sort(self.source.get_graph())
2532
version_list = [i for i in full_list if (not self.target.has_version(i)
2533
and i in needed_versions)]
2537
copy_queue_records = []
2539
for version_id in version_list:
2540
options = self.source._index.get_options(version_id)
2541
parents = self.source._index.get_parents_with_ghosts(version_id)
2542
# check that its will be a consistent copy:
2543
for parent in parents:
2544
# if source has the parent, we must :
2545
# * already have it or
2546
# * have it scheduled already
2547
# otherwise we don't care
2548
assert (self.target.has_version(parent) or
2549
parent in copy_set or
2550
not self.source.has_version(parent))
2551
index_memo = self.source._index.get_position(version_id)
2552
copy_queue_records.append((version_id, index_memo))
2553
copy_queue.append((version_id, options, parents))
2554
copy_set.add(version_id)
2556
# data suck the join:
2558
total = len(version_list)
2561
for (version_id, raw_data), \
2562
(version_id2, options, parents) in \
2563
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2565
assert version_id == version_id2, 'logic error, inconsistent results'
2567
pb.update("Joining knit", count, total)
2569
size, raw_data = converter(raw_data, version_id, options,
2572
size = len(raw_data)
2573
raw_records.append((version_id, options, parents, size))
2574
raw_datum.append(raw_data)
2575
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2580
def _anno_to_plain_converter(self, raw_data, version_id, options,
2582
"""Convert annotated content to plain content."""
2583
data, digest = self.source._data._parse_record(version_id, raw_data)
2584
if 'fulltext' in options:
2585
content = self.source.factory.parse_fulltext(data, version_id)
2586
lines = self.target.factory.lower_fulltext(content)
2588
delta = self.source.factory.parse_line_delta(data, version_id,
2590
lines = self.target.factory.lower_line_delta(delta)
2591
return self.target._data._record_to_data(version_id, digest, lines)
2594
InterVersionedFile.register_optimiser(InterKnit)
2597
class WeaveToKnit(InterVersionedFile):
2598
"""Optimised code paths for weave to knit operations."""
2600
_matching_file_from_factory = bzrlib.weave.WeaveFile
2601
_matching_file_to_factory = KnitVersionedFile
2604
def is_compatible(source, target):
2605
"""Be compatible with weaves to knits."""
2607
return (isinstance(source, bzrlib.weave.Weave) and
2608
isinstance(target, KnitVersionedFile))
2609
except AttributeError:
2612
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2613
"""See InterVersionedFile.join."""
2614
assert isinstance(self.source, bzrlib.weave.Weave)
2615
assert isinstance(self.target, KnitVersionedFile)
2617
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2622
pb = ui.ui_factory.nested_progress_bar()
2624
version_ids = list(version_ids)
2626
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2627
this_versions = set(self.target._index.get_versions())
2628
needed_versions = self.source_ancestry - this_versions
2630
if not needed_versions:
2632
full_list = topo_sort(self.source.get_graph())
2634
version_list = [i for i in full_list if (not self.target.has_version(i)
2635
and i in needed_versions)]
2639
total = len(version_list)
2640
for version_id in version_list:
2641
pb.update("Converting to knit", count, total)
2642
parents = self.source.get_parents(version_id)
2643
# check that its will be a consistent copy:
2644
for parent in parents:
2645
# if source has the parent, we must already have it
2646
assert (self.target.has_version(parent))
2647
self.target.add_lines(
2648
version_id, parents, self.source.get_lines(version_id))
2655
InterVersionedFile.register_optimiser(WeaveToKnit)
2658
# Deprecated, use PatienceSequenceMatcher instead
2659
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2662
def annotate_knit(knit, revision_id):
2663
"""Annotate a knit with no cached annotations.
2665
This implementation is for knits with no cached annotations.
2666
It will work for knits with cached annotations, but this is not
2669
ancestry = knit.get_ancestry(revision_id)
2670
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2672
for candidate in ancestry:
2673
if candidate in annotations:
2675
parents = knit.get_parents(candidate)
2676
if len(parents) == 0:
2678
elif knit._index.get_method(candidate) != 'line-delta':
2681
parent, sha1, noeol, delta = knit.get_delta(candidate)
2682
blocks = KnitContent.get_line_delta_blocks(delta,
2683
fulltext[parents[0]], fulltext[candidate])
2684
annotations[candidate] = list(annotate.reannotate([annotations[p]
2685
for p in parents], fulltext[candidate], candidate, blocks))
2686
return iter(annotations[revision_id])
2690
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2692
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
