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
if 'knit' in debug.debug_flags:
741
'incompatible format signature inserting to %r', self)
742
source = self._knit_from_datastream(
743
(format, data_list, reader_callable))
747
for version_id, options, length, parents in data_list:
748
if self.has_version(version_id):
749
# First check: the list of parents.
750
my_parents = self.get_parents_with_ghosts(version_id)
751
if tuple(my_parents) != tuple(parents):
752
# XXX: KnitCorrupt is not quite the right exception here.
755
'parents list %r from data stream does not match '
756
'already recorded parents %r for %s'
757
% (parents, my_parents, version_id))
759
# Also check the SHA-1 of the fulltext this content will
761
raw_data = reader_callable(length)
762
my_fulltext_sha1 = self.get_sha1(version_id)
763
df, rec = self._data._parse_record_header(version_id, raw_data)
764
stream_fulltext_sha1 = rec[3]
765
if my_fulltext_sha1 != stream_fulltext_sha1:
766
# Actually, we don't know if it's this knit that's corrupt,
767
# or the data stream we're trying to insert.
769
self.filename, 'sha-1 does not match %s' % version_id)
771
if 'line-delta' in options:
772
# Make sure that this knit record is actually useful: a
773
# line-delta is no use unless we have its parent.
774
# Fetching from a broken repository with this problem
775
# shouldn't break the target repository.
777
# See https://bugs.launchpad.net/bzr/+bug/164443
778
if not self._index.has_version(parents[0]):
781
'line-delta from stream '
784
'missing parent %s\n'
785
'Try running "bzr check" '
786
'on the source repository, and "bzr reconcile" '
788
(version_id, parents[0]))
789
self._add_raw_records(
790
[(version_id, options, parents, length)],
791
reader_callable(length))
793
def _knit_from_datastream(self, (format, data_list, reader_callable)):
794
"""Create a knit object from a data stream.
796
This method exists to allow conversion of data streams that do not
797
match the signature of this knit. Generally it will be slower and use
798
more memory to use this method to insert data, but it will work.
800
:seealso: get_data_stream for details on datastreams.
801
:return: A knit versioned file which can be used to join the datastream
804
if format == "knit-plain":
805
factory = KnitPlainFactory()
806
elif format == "knit-annotated":
807
factory = KnitAnnotateFactory()
809
raise errors.KnitDataStreamUnknown(format)
810
index = _StreamIndex(data_list)
811
access = _StreamAccess(reader_callable, index, self, factory)
812
return KnitVersionedFile(self.filename, self.transport,
813
factory=factory, index=index, access_method=access)
816
"""See VersionedFile.versions."""
817
if 'evil' in debug.debug_flags:
818
trace.mutter_callsite(2, "versions scales with size of history")
819
return self._index.get_versions()
821
def has_version(self, version_id):
822
"""See VersionedFile.has_version."""
823
if 'evil' in debug.debug_flags:
824
trace.mutter_callsite(2, "has_version is a LBYL scenario")
825
return self._index.has_version(version_id)
827
__contains__ = has_version
829
def _merge_annotations(self, content, parents, parent_texts={},
830
delta=None, annotated=None,
831
left_matching_blocks=None):
832
"""Merge annotations for content. This is done by comparing
833
the annotations based on changed to the text.
835
if left_matching_blocks is not None:
836
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
840
for parent_id in parents:
841
merge_content = self._get_content(parent_id, parent_texts)
842
if (parent_id == parents[0] and delta_seq is not None):
845
seq = patiencediff.PatienceSequenceMatcher(
846
None, merge_content.text(), content.text())
847
for i, j, n in seq.get_matching_blocks():
850
# this appears to copy (origin, text) pairs across to the
851
# new content for any line that matches the last-checked
853
content._lines[j:j+n] = merge_content._lines[i:i+n]
855
if delta_seq is None:
856
reference_content = self._get_content(parents[0], parent_texts)
857
new_texts = content.text()
858
old_texts = reference_content.text()
859
delta_seq = patiencediff.PatienceSequenceMatcher(
860
None, old_texts, new_texts)
861
return self._make_line_delta(delta_seq, content)
863
def _make_line_delta(self, delta_seq, new_content):
864
"""Generate a line delta from delta_seq and new_content."""
866
for op in delta_seq.get_opcodes():
869
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
872
def _get_components_positions(self, version_ids):
873
"""Produce a map of position data for the components of versions.
875
This data is intended to be used for retrieving the knit records.
877
A dict of version_id to (method, data_pos, data_size, next) is
879
method is the way referenced data should be applied.
880
data_pos is the position of the data in the knit.
881
data_size is the size of the data in the knit.
882
next is the build-parent of the version, or None for fulltexts.
885
for version_id in version_ids:
888
while cursor is not None and cursor not in component_data:
889
method = self._index.get_method(cursor)
890
if method == 'fulltext':
893
next = self.get_parents_with_ghosts(cursor)[0]
894
index_memo = self._index.get_position(cursor)
895
component_data[cursor] = (method, index_memo, next)
897
return component_data
899
def _get_content(self, version_id, parent_texts={}):
900
"""Returns a content object that makes up the specified
902
cached_version = parent_texts.get(version_id, None)
903
if cached_version is not None:
904
if not self.has_version(version_id):
905
raise RevisionNotPresent(version_id, self.filename)
906
return cached_version
908
text_map, contents_map = self._get_content_maps([version_id])
909
return contents_map[version_id]
911
def _check_versions_present(self, version_ids):
912
"""Check that all specified versions are present."""
913
self._index.check_versions_present(version_ids)
915
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
916
nostore_sha, random_id, check_content):
917
"""See VersionedFile.add_lines_with_ghosts()."""
918
self._check_add(version_id, lines, random_id, check_content)
919
return self._add(version_id, lines, parents, self.delta,
920
parent_texts, None, nostore_sha, random_id)
922
def _add_lines(self, version_id, parents, lines, parent_texts,
923
left_matching_blocks, nostore_sha, random_id, check_content):
924
"""See VersionedFile.add_lines."""
925
self._check_add(version_id, lines, random_id, check_content)
926
self._check_versions_present(parents)
927
return self._add(version_id, lines[:], parents, self.delta,
928
parent_texts, left_matching_blocks, nostore_sha, random_id)
930
def _check_add(self, version_id, lines, random_id, check_content):
931
"""check that version_id and lines are safe to add."""
932
if contains_whitespace(version_id):
933
raise InvalidRevisionId(version_id, self.filename)
934
self.check_not_reserved_id(version_id)
935
# Technically this could be avoided if we are happy to allow duplicate
936
# id insertion when other things than bzr core insert texts, but it
937
# seems useful for folk using the knit api directly to have some safety
938
# blanket that we can disable.
939
if not random_id and self.has_version(version_id):
940
raise RevisionAlreadyPresent(version_id, self.filename)
942
self._check_lines_not_unicode(lines)
943
self._check_lines_are_lines(lines)
945
def _add(self, version_id, lines, parents, delta, parent_texts,
946
left_matching_blocks, nostore_sha, random_id):
947
"""Add a set of lines on top of version specified by parents.
949
If delta is true, compress the text as a line-delta against
952
Any versions not present will be converted into ghosts.
954
# first thing, if the content is something we don't need to store, find
956
line_bytes = ''.join(lines)
957
digest = sha_string(line_bytes)
958
if nostore_sha == digest:
959
raise errors.ExistingContent
962
if parent_texts is None:
964
for parent in parents:
965
if self.has_version(parent):
966
present_parents.append(parent)
968
# can only compress against the left most present parent.
970
(len(present_parents) == 0 or
971
present_parents[0] != parents[0])):
974
text_length = len(line_bytes)
977
if lines[-1][-1] != '\n':
978
# copy the contents of lines.
980
options.append('no-eol')
981
lines[-1] = lines[-1] + '\n'
985
# To speed the extract of texts the delta chain is limited
986
# to a fixed number of deltas. This should minimize both
987
# I/O and the time spend applying deltas.
988
delta = self._check_should_delta(present_parents)
990
assert isinstance(version_id, str)
991
content = self.factory.make(lines, version_id)
992
if delta or (self.factory.annotated and len(present_parents) > 0):
993
# Merge annotations from parent texts if needed.
994
delta_hunks = self._merge_annotations(content, present_parents,
995
parent_texts, delta, self.factory.annotated,
996
left_matching_blocks)
999
options.append('line-delta')
1000
store_lines = self.factory.lower_line_delta(delta_hunks)
1001
size, bytes = self._data._record_to_data(version_id, digest,
1004
options.append('fulltext')
1005
# isinstance is slower and we have no hierarchy.
1006
if self.factory.__class__ == KnitPlainFactory:
1007
# Use the already joined bytes saving iteration time in
1009
size, bytes = self._data._record_to_data(version_id, digest,
1010
lines, [line_bytes])
1012
# get mixed annotation + content and feed it into the
1014
store_lines = self.factory.lower_fulltext(content)
1015
size, bytes = self._data._record_to_data(version_id, digest,
1018
access_memo = self._data.add_raw_records([size], bytes)[0]
1019
self._index.add_versions(
1020
((version_id, options, access_memo, parents),),
1021
random_id=random_id)
1022
return digest, text_length, content
1024
def check(self, progress_bar=None):
1025
"""See VersionedFile.check()."""
1027
def _clone_text(self, new_version_id, old_version_id, parents):
1028
"""See VersionedFile.clone_text()."""
1029
# FIXME RBC 20060228 make fast by only inserting an index with null
1031
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1033
def get_lines(self, version_id):
1034
"""See VersionedFile.get_lines()."""
1035
return self.get_line_list([version_id])[0]
1037
def _get_record_map(self, version_ids):
1038
"""Produce a dictionary of knit records.
1040
The keys are version_ids, the values are tuples of (method, content,
1042
method is the way the content should be applied.
1043
content is a KnitContent object.
1044
digest is the SHA1 digest of this version id after all steps are done
1045
next is the build-parent of the version, i.e. the leftmost ancestor.
1046
If the method is fulltext, next will be None.
1048
position_map = self._get_components_positions(version_ids)
1049
# c = component_id, m = method, i_m = index_memo, n = next
1050
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1052
for component_id, content, digest in \
1053
self._data.read_records_iter(records):
1054
method, index_memo, next = position_map[component_id]
1055
record_map[component_id] = method, content, digest, next
1059
def get_text(self, version_id):
1060
"""See VersionedFile.get_text"""
1061
return self.get_texts([version_id])[0]
1063
def get_texts(self, version_ids):
1064
return [''.join(l) for l in self.get_line_list(version_ids)]
1066
def get_line_list(self, version_ids):
1067
"""Return the texts of listed versions as a list of strings."""
1068
for version_id in version_ids:
1069
self.check_not_reserved_id(version_id)
1070
text_map, content_map = self._get_content_maps(version_ids)
1071
return [text_map[v] for v in version_ids]
1073
_get_lf_split_line_list = get_line_list
1075
def _get_content_maps(self, version_ids):
1076
"""Produce maps of text and KnitContents
1078
:return: (text_map, content_map) where text_map contains the texts for
1079
the requested versions and content_map contains the KnitContents.
1080
Both dicts take version_ids as their keys.
1082
# FUTURE: This function could be improved for the 'extract many' case
1083
# by tracking each component and only doing the copy when the number of
1084
# children than need to apply delta's to it is > 1 or it is part of the
1086
version_ids = list(version_ids)
1087
multiple_versions = len(version_ids) != 1
1088
record_map = self._get_record_map(version_ids)
1093
for version_id in version_ids:
1096
while cursor is not None:
1097
method, data, digest, next = record_map[cursor]
1098
components.append((cursor, method, data, digest))
1099
if cursor in content_map:
1104
for component_id, method, data, digest in reversed(components):
1105
if component_id in content_map:
1106
content = content_map[component_id]
1108
if method == 'fulltext':
1109
assert content is None
1110
content = self.factory.parse_fulltext(data, version_id)
1111
elif method == 'line-delta':
1112
delta = self.factory.parse_line_delta(data, version_id)
1113
if multiple_versions:
1114
# only doing this when we want multiple versions
1115
# output avoids list copies - which reference and
1116
# dereference many strings.
1117
content = content.copy()
1118
content.apply_delta(delta, version_id)
1119
if multiple_versions:
1120
content_map[component_id] = content
1122
if 'no-eol' in self._index.get_options(version_id):
1123
if multiple_versions:
1124
content = content.copy()
1125
content.strip_last_line_newline()
1126
final_content[version_id] = content
1128
# digest here is the digest from the last applied component.
1129
text = content.text()
1130
actual_sha = sha_strings(text)
1131
if actual_sha != digest:
1132
raise KnitCorrupt(self.filename,
1134
'\n of reconstructed text does not match'
1136
'\n for version %s' %
1137
(actual_sha, digest, version_id))
1138
text_map[version_id] = text
1139
return text_map, final_content
1141
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1143
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1144
if version_ids is None:
1145
version_ids = self.versions()
1147
pb = progress.DummyProgress()
1148
# we don't care about inclusions, the caller cares.
1149
# but we need to setup a list of records to visit.
1150
# we need version_id, position, length
1151
version_id_records = []
1152
requested_versions = set(version_ids)
1153
# filter for available versions
1154
for version_id in requested_versions:
1155
if not self.has_version(version_id):
1156
raise RevisionNotPresent(version_id, self.filename)
1157
# get a in-component-order queue:
1158
for version_id in self.versions():
1159
if version_id in requested_versions:
1160
index_memo = self._index.get_position(version_id)
1161
version_id_records.append((version_id, index_memo))
1163
total = len(version_id_records)
1164
for version_idx, (version_id, data, sha_value) in \
1165
enumerate(self._data.read_records_iter(version_id_records)):
1166
pb.update('Walking content.', version_idx, total)
1167
method = self._index.get_method(version_id)
1169
assert method in ('fulltext', 'line-delta')
1170
if method == 'fulltext':
1171
line_iterator = self.factory.get_fulltext_content(data)
1173
line_iterator = self.factory.get_linedelta_content(data)
1174
# XXX: It might be more efficient to yield (version_id,
1175
# line_iterator) in the future. However for now, this is a simpler
1176
# change to integrate into the rest of the codebase. RBC 20071110
1177
for line in line_iterator:
1178
yield line, version_id
1180
pb.update('Walking content.', total, total)
1182
def iter_parents(self, version_ids):
1183
"""Iterate through the parents for many version ids.
1185
:param version_ids: An iterable yielding version_ids.
1186
:return: An iterator that yields (version_id, parents). Requested
1187
version_ids not present in the versioned file are simply skipped.
1188
The order is undefined, allowing for different optimisations in
1189
the underlying implementation.
1191
return self._index.iter_parents(version_ids)
1193
def num_versions(self):
1194
"""See VersionedFile.num_versions()."""
1195
return self._index.num_versions()
1197
__len__ = num_versions
1199
def annotate_iter(self, version_id):
1200
"""See VersionedFile.annotate_iter."""
1201
return self.factory.annotate_iter(self, version_id)
1203
def get_parents(self, version_id):
1204
"""See VersionedFile.get_parents."""
1207
# 52554 calls in 1264 872 internal down from 3674
1209
return self._index.get_parents(version_id)
1211
raise RevisionNotPresent(version_id, self.filename)
1213
def get_parents_with_ghosts(self, version_id):
1214
"""See VersionedFile.get_parents."""
1216
return self._index.get_parents_with_ghosts(version_id)
1218
raise RevisionNotPresent(version_id, self.filename)
1220
def get_ancestry(self, versions, topo_sorted=True):
1221
"""See VersionedFile.get_ancestry."""
1222
if isinstance(versions, basestring):
1223
versions = [versions]
1226
return self._index.get_ancestry(versions, topo_sorted)
1228
def get_ancestry_with_ghosts(self, versions):
1229
"""See VersionedFile.get_ancestry_with_ghosts."""
1230
if isinstance(versions, basestring):
1231
versions = [versions]
1234
return self._index.get_ancestry_with_ghosts(versions)
1236
def plan_merge(self, ver_a, ver_b):
1237
"""See VersionedFile.plan_merge."""
1238
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1239
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1240
annotated_a = self.annotate(ver_a)
1241
annotated_b = self.annotate(ver_b)
1242
return merge._plan_annotate_merge(annotated_a, annotated_b,
1243
ancestors_a, ancestors_b)
1246
class _KnitComponentFile(object):
1247
"""One of the files used to implement a knit database"""
1249
def __init__(self, transport, filename, mode, file_mode=None,
1250
create_parent_dir=False, dir_mode=None):
1251
self._transport = transport
1252
self._filename = filename
1254
self._file_mode = file_mode
1255
self._dir_mode = dir_mode
1256
self._create_parent_dir = create_parent_dir
1257
self._need_to_create = False
1259
def _full_path(self):
1260
"""Return the full path to this file."""
1261
return self._transport.base + self._filename
1263
def check_header(self, fp):
1264
line = fp.readline()
1266
# An empty file can actually be treated as though the file doesn't
1268
raise errors.NoSuchFile(self._full_path())
1269
if line != self.HEADER:
1270
raise KnitHeaderError(badline=line,
1271
filename=self._transport.abspath(self._filename))
1274
return '%s(%s)' % (self.__class__.__name__, self._filename)
1277
class _KnitIndex(_KnitComponentFile):
1278
"""Manages knit index file.
1280
The index is already kept in memory and read on startup, to enable
1281
fast lookups of revision information. The cursor of the index
1282
file is always pointing to the end, making it easy to append
1285
_cache is a cache for fast mapping from version id to a Index
1288
_history is a cache for fast mapping from indexes to version ids.
1290
The index data format is dictionary compressed when it comes to
1291
parent references; a index entry may only have parents that with a
1292
lover index number. As a result, the index is topological sorted.
1294
Duplicate entries may be written to the index for a single version id
1295
if this is done then the latter one completely replaces the former:
1296
this allows updates to correct version and parent information.
1297
Note that the two entries may share the delta, and that successive
1298
annotations and references MUST point to the first entry.
1300
The index file on disc contains a header, followed by one line per knit
1301
record. The same revision can be present in an index file more than once.
1302
The first occurrence gets assigned a sequence number starting from 0.
1304
The format of a single line is
1305
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1306
REVISION_ID is a utf8-encoded revision id
1307
FLAGS is a comma separated list of flags about the record. Values include
1308
no-eol, line-delta, fulltext.
1309
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1310
that the the compressed data starts at.
1311
LENGTH is the ascii representation of the length of the data file.
1312
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1314
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1315
revision id already in the knit that is a parent of REVISION_ID.
1316
The ' :' marker is the end of record marker.
1319
when a write is interrupted to the index file, it will result in a line
1320
that does not end in ' :'. If the ' :' is not present at the end of a line,
1321
or at the end of the file, then the record that is missing it will be
1322
ignored by the parser.
1324
When writing new records to the index file, the data is preceded by '\n'
1325
to ensure that records always start on new lines even if the last write was
1326
interrupted. As a result its normal for the last line in the index to be
1327
missing a trailing newline. One can be added with no harmful effects.
1330
HEADER = "# bzr knit index 8\n"
1332
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1333
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1335
def _cache_version(self, version_id, options, pos, size, parents):
1336
"""Cache a version record in the history array and index cache.
1338
This is inlined into _load_data for performance. KEEP IN SYNC.
1339
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1342
# only want the _history index to reference the 1st index entry
1344
if version_id not in self._cache:
1345
index = len(self._history)
1346
self._history.append(version_id)
1348
index = self._cache[version_id][5]
1349
self._cache[version_id] = (version_id,
1356
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1357
create_parent_dir=False, delay_create=False, dir_mode=None):
1358
_KnitComponentFile.__init__(self, transport, filename, mode,
1359
file_mode=file_mode,
1360
create_parent_dir=create_parent_dir,
1363
# position in _history is the 'official' index for a revision
1364
# but the values may have come from a newer entry.
1365
# so - wc -l of a knit index is != the number of unique names
1369
fp = self._transport.get(self._filename)
1371
# _load_data may raise NoSuchFile if the target knit is
1373
_load_data(self, fp)
1377
if mode != 'w' or not create:
1380
self._need_to_create = True
1382
self._transport.put_bytes_non_atomic(
1383
self._filename, self.HEADER, mode=self._file_mode)
1385
def get_graph(self):
1386
"""Return a list of the node:parents lists from this knit index."""
1387
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1389
def get_ancestry(self, versions, topo_sorted=True):
1390
"""See VersionedFile.get_ancestry."""
1391
# get a graph of all the mentioned versions:
1393
pending = set(versions)
1396
version = pending.pop()
1399
parents = [p for p in cache[version][4] if p in cache]
1401
raise RevisionNotPresent(version, self._filename)
1402
# if not completed and not a ghost
1403
pending.update([p for p in parents if p not in graph])
1404
graph[version] = parents
1407
return topo_sort(graph.items())
1409
def get_ancestry_with_ghosts(self, versions):
1410
"""See VersionedFile.get_ancestry_with_ghosts."""
1411
# get a graph of all the mentioned versions:
1412
self.check_versions_present(versions)
1415
pending = set(versions)
1417
version = pending.pop()
1419
parents = cache[version][4]
1425
pending.update([p for p in parents if p not in graph])
1426
graph[version] = parents
1427
return topo_sort(graph.items())
1429
def iter_parents(self, version_ids):
1430
"""Iterate through the parents for many version ids.
1432
:param version_ids: An iterable yielding version_ids.
1433
:return: An iterator that yields (version_id, parents). Requested
1434
version_ids not present in the versioned file are simply skipped.
1435
The order is undefined, allowing for different optimisations in
1436
the underlying implementation.
1438
for version_id in version_ids:
1440
yield version_id, tuple(self.get_parents(version_id))
1444
def num_versions(self):
1445
return len(self._history)
1447
__len__ = num_versions
1449
def get_versions(self):
1450
"""Get all the versions in the file. not topologically sorted."""
1451
return self._history
1453
def _version_list_to_index(self, versions):
1456
for version in versions:
1457
if version in cache:
1458
# -- inlined lookup() --
1459
result_list.append(str(cache[version][5]))
1460
# -- end lookup () --
1462
result_list.append('.' + version)
1463
return ' '.join(result_list)
1465
def add_version(self, version_id, options, index_memo, parents):
1466
"""Add a version record to the index."""
1467
self.add_versions(((version_id, options, index_memo, parents),))
1469
def add_versions(self, versions, random_id=False):
1470
"""Add multiple versions to the index.
1472
:param versions: a list of tuples:
1473
(version_id, options, pos, size, parents).
1474
:param random_id: If True the ids being added were randomly generated
1475
and no check for existence will be performed.
1478
orig_history = self._history[:]
1479
orig_cache = self._cache.copy()
1482
for version_id, options, (index, pos, size), parents in versions:
1483
line = "\n%s %s %s %s %s :" % (version_id,
1487
self._version_list_to_index(parents))
1488
assert isinstance(line, str), \
1489
'content must be utf-8 encoded: %r' % (line,)
1491
self._cache_version(version_id, options, pos, size, parents)
1492
if not self._need_to_create:
1493
self._transport.append_bytes(self._filename, ''.join(lines))
1496
sio.write(self.HEADER)
1497
sio.writelines(lines)
1499
self._transport.put_file_non_atomic(self._filename, sio,
1500
create_parent_dir=self._create_parent_dir,
1501
mode=self._file_mode,
1502
dir_mode=self._dir_mode)
1503
self._need_to_create = False
1505
# If any problems happen, restore the original values and re-raise
1506
self._history = orig_history
1507
self._cache = orig_cache
1510
def has_version(self, version_id):
1511
"""True if the version is in the index."""
1512
return version_id in self._cache
1514
def get_position(self, version_id):
1515
"""Return details needed to access the version.
1517
.kndx indices do not support split-out data, so return None for the
1520
:return: a tuple (None, data position, size) to hand to the access
1521
logic to get the record.
1523
entry = self._cache[version_id]
1524
return None, entry[2], entry[3]
1526
def get_method(self, version_id):
1527
"""Return compression method of specified version."""
1529
options = self._cache[version_id][1]
1531
raise RevisionNotPresent(version_id, self._filename)
1532
if 'fulltext' in options:
1535
if 'line-delta' not in options:
1536
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1539
def get_options(self, version_id):
1540
"""Return a list representing options.
1544
return self._cache[version_id][1]
1546
def get_parents(self, version_id):
1547
"""Return parents of specified version ignoring ghosts."""
1548
return [parent for parent in self._cache[version_id][4]
1549
if parent in self._cache]
1551
def get_parents_with_ghosts(self, version_id):
1552
"""Return parents of specified version with ghosts."""
1553
return self._cache[version_id][4]
1555
def check_versions_present(self, version_ids):
1556
"""Check that all specified versions are present."""
1558
for version_id in version_ids:
1559
if version_id not in cache:
1560
raise RevisionNotPresent(version_id, self._filename)
1563
class KnitGraphIndex(object):
1564
"""A knit index that builds on GraphIndex."""
1566
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1567
"""Construct a KnitGraphIndex on a graph_index.
1569
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1570
:param deltas: Allow delta-compressed records.
1571
:param add_callback: If not None, allow additions to the index and call
1572
this callback with a list of added GraphIndex nodes:
1573
[(node, value, node_refs), ...]
1574
:param parents: If True, record knits parents, if not do not record
1577
self._graph_index = graph_index
1578
self._deltas = deltas
1579
self._add_callback = add_callback
1580
self._parents = parents
1581
if deltas and not parents:
1582
raise KnitCorrupt(self, "Cannot do delta compression without "
1585
def _get_entries(self, keys, check_present=False):
1586
"""Get the entries for keys.
1588
:param keys: An iterable of index keys, - 1-tuples.
1593
for node in self._graph_index.iter_entries(keys):
1595
found_keys.add(node[1])
1597
# adapt parentless index to the rest of the code.
1598
for node in self._graph_index.iter_entries(keys):
1599
yield node[0], node[1], node[2], ()
1600
found_keys.add(node[1])
1602
missing_keys = keys.difference(found_keys)
1604
raise RevisionNotPresent(missing_keys.pop(), self)
1606
def _present_keys(self, version_ids):
1608
node[1] for node in self._get_entries(version_ids)])
1610
def _parentless_ancestry(self, versions):
1611
"""Honour the get_ancestry API for parentless knit indices."""
1612
wanted_keys = self._version_ids_to_keys(versions)
1613
present_keys = self._present_keys(wanted_keys)
1614
missing = set(wanted_keys).difference(present_keys)
1616
raise RevisionNotPresent(missing.pop(), self)
1617
return list(self._keys_to_version_ids(present_keys))
1619
def get_ancestry(self, versions, topo_sorted=True):
1620
"""See VersionedFile.get_ancestry."""
1621
if not self._parents:
1622
return self._parentless_ancestry(versions)
1623
# XXX: This will do len(history) index calls - perhaps
1624
# it should be altered to be a index core feature?
1625
# get a graph of all the mentioned versions:
1628
versions = self._version_ids_to_keys(versions)
1629
pending = set(versions)
1631
# get all pending nodes
1632
this_iteration = pending
1633
new_nodes = self._get_entries(this_iteration)
1636
for (index, key, value, node_refs) in new_nodes:
1637
# dont ask for ghosties - otherwise
1638
# we we can end up looping with pending
1639
# being entirely ghosted.
1640
graph[key] = [parent for parent in node_refs[0]
1641
if parent not in ghosts]
1643
for parent in graph[key]:
1644
# dont examine known nodes again
1649
ghosts.update(this_iteration.difference(found))
1650
if versions.difference(graph):
1651
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1653
result_keys = topo_sort(graph.items())
1655
result_keys = graph.iterkeys()
1656
return [key[0] for key in result_keys]
1658
def get_ancestry_with_ghosts(self, versions):
1659
"""See VersionedFile.get_ancestry."""
1660
if not self._parents:
1661
return self._parentless_ancestry(versions)
1662
# XXX: This will do len(history) index calls - perhaps
1663
# it should be altered to be a index core feature?
1664
# get a graph of all the mentioned versions:
1666
versions = self._version_ids_to_keys(versions)
1667
pending = set(versions)
1669
# get all pending nodes
1670
this_iteration = pending
1671
new_nodes = self._get_entries(this_iteration)
1673
for (index, key, value, node_refs) in new_nodes:
1674
graph[key] = node_refs[0]
1676
for parent in graph[key]:
1677
# dont examine known nodes again
1681
missing_versions = this_iteration.difference(graph)
1682
missing_needed = versions.intersection(missing_versions)
1684
raise RevisionNotPresent(missing_needed.pop(), self)
1685
for missing_version in missing_versions:
1686
# add a key, no parents
1687
graph[missing_version] = []
1688
pending.discard(missing_version) # don't look for it
1689
result_keys = topo_sort(graph.items())
1690
return [key[0] for key in result_keys]
1692
def get_graph(self):
1693
"""Return a list of the node:parents lists from this knit index."""
1694
if not self._parents:
1695
return [(key, ()) for key in self.get_versions()]
1697
for index, key, value, refs in self._graph_index.iter_all_entries():
1698
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1701
def iter_parents(self, version_ids):
1702
"""Iterate through the parents for many version ids.
1704
:param version_ids: An iterable yielding version_ids.
1705
:return: An iterator that yields (version_id, parents). Requested
1706
version_ids not present in the versioned file are simply skipped.
1707
The order is undefined, allowing for different optimisations in
1708
the underlying implementation.
1711
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1713
present_parents = set()
1714
for node in all_nodes:
1715
all_parents.update(node[3][0])
1716
# any node we are querying must be present
1717
present_parents.add(node[1])
1718
unknown_parents = all_parents.difference(present_parents)
1719
present_parents.update(self._present_keys(unknown_parents))
1720
for node in all_nodes:
1722
for parent in node[3][0]:
1723
if parent in present_parents:
1724
parents.append(parent[0])
1725
yield node[1][0], tuple(parents)
1727
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1728
yield node[1][0], ()
1730
def num_versions(self):
1731
return len(list(self._graph_index.iter_all_entries()))
1733
__len__ = num_versions
1735
def get_versions(self):
1736
"""Get all the versions in the file. not topologically sorted."""
1737
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1739
def has_version(self, version_id):
1740
"""True if the version is in the index."""
1741
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1743
def _keys_to_version_ids(self, keys):
1744
return tuple(key[0] for key in keys)
1746
def get_position(self, version_id):
1747
"""Return details needed to access the version.
1749
:return: a tuple (index, data position, size) to hand to the access
1750
logic to get the record.
1752
node = self._get_node(version_id)
1753
bits = node[2][1:].split(' ')
1754
return node[0], int(bits[0]), int(bits[1])
1756
def get_method(self, version_id):
1757
"""Return compression method of specified version."""
1758
if not self._deltas:
1760
return self._parent_compression(self._get_node(version_id)[3][1])
1762
def _parent_compression(self, reference_list):
1763
# use the second reference list to decide if this is delta'd or not.
1764
if len(reference_list):
1769
def _get_node(self, version_id):
1771
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1773
raise RevisionNotPresent(version_id, self)
1775
def get_options(self, version_id):
1776
"""Return a list representing options.
1780
node = self._get_node(version_id)
1781
if not self._deltas:
1782
options = ['fulltext']
1784
options = [self._parent_compression(node[3][1])]
1785
if node[2][0] == 'N':
1786
options.append('no-eol')
1789
def get_parents(self, version_id):
1790
"""Return parents of specified version ignoring ghosts."""
1791
parents = list(self.iter_parents([version_id]))
1794
raise errors.RevisionNotPresent(version_id, self)
1795
return parents[0][1]
1797
def get_parents_with_ghosts(self, version_id):
1798
"""Return parents of specified version with ghosts."""
1799
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1800
check_present=True))
1801
if not self._parents:
1803
return self._keys_to_version_ids(nodes[0][3][0])
1805
def check_versions_present(self, version_ids):
1806
"""Check that all specified versions are present."""
1807
keys = self._version_ids_to_keys(version_ids)
1808
present = self._present_keys(keys)
1809
missing = keys.difference(present)
1811
raise RevisionNotPresent(missing.pop(), self)
1813
def add_version(self, version_id, options, access_memo, parents):
1814
"""Add a version record to the index."""
1815
return self.add_versions(((version_id, options, access_memo, parents),))
1817
def add_versions(self, versions, random_id=False):
1818
"""Add multiple versions to the index.
1820
This function does not insert data into the Immutable GraphIndex
1821
backing the KnitGraphIndex, instead it prepares data for insertion by
1822
the caller and checks that it is safe to insert then calls
1823
self._add_callback with the prepared GraphIndex nodes.
1825
:param versions: a list of tuples:
1826
(version_id, options, pos, size, parents).
1827
:param random_id: If True the ids being added were randomly generated
1828
and no check for existence will be performed.
1830
if not self._add_callback:
1831
raise errors.ReadOnlyError(self)
1832
# we hope there are no repositories with inconsistent parentage
1837
for (version_id, options, access_memo, parents) in versions:
1838
index, pos, size = access_memo
1839
key = (version_id, )
1840
parents = tuple((parent, ) for parent in parents)
1841
if 'no-eol' in options:
1845
value += "%d %d" % (pos, size)
1846
if not self._deltas:
1847
if 'line-delta' in options:
1848
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1851
if 'line-delta' in options:
1852
node_refs = (parents, (parents[0],))
1854
node_refs = (parents, ())
1856
node_refs = (parents, )
1859
raise KnitCorrupt(self, "attempt to add node with parents "
1860
"in parentless index.")
1862
keys[key] = (value, node_refs)
1864
present_nodes = self._get_entries(keys)
1865
for (index, key, value, node_refs) in present_nodes:
1866
if (value, node_refs) != keys[key]:
1867
raise KnitCorrupt(self, "inconsistent details in add_versions"
1868
": %s %s" % ((value, node_refs), keys[key]))
1872
for key, (value, node_refs) in keys.iteritems():
1873
result.append((key, value, node_refs))
1875
for key, (value, node_refs) in keys.iteritems():
1876
result.append((key, value))
1877
self._add_callback(result)
1879
def _version_ids_to_keys(self, version_ids):
1880
return set((version_id, ) for version_id in version_ids)
1883
class _KnitAccess(object):
1884
"""Access to knit records in a .knit file."""
1886
def __init__(self, transport, filename, _file_mode, _dir_mode,
1887
_need_to_create, _create_parent_dir):
1888
"""Create a _KnitAccess for accessing and inserting data.
1890
:param transport: The transport the .knit is located on.
1891
:param filename: The filename of the .knit.
1893
self._transport = transport
1894
self._filename = filename
1895
self._file_mode = _file_mode
1896
self._dir_mode = _dir_mode
1897
self._need_to_create = _need_to_create
1898
self._create_parent_dir = _create_parent_dir
1900
def add_raw_records(self, sizes, raw_data):
1901
"""Add raw knit bytes to a storage area.
1903
The data is spooled to whereever the access method is storing data.
1905
:param sizes: An iterable containing the size of each raw data segment.
1906
:param raw_data: A bytestring containing the data.
1907
:return: A list of memos to retrieve the record later. Each memo is a
1908
tuple - (index, pos, length), where the index field is always None
1909
for the .knit access method.
1911
assert type(raw_data) == str, \
1912
'data must be plain bytes was %s' % type(raw_data)
1913
if not self._need_to_create:
1914
base = self._transport.append_bytes(self._filename, raw_data)
1916
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1917
create_parent_dir=self._create_parent_dir,
1918
mode=self._file_mode,
1919
dir_mode=self._dir_mode)
1920
self._need_to_create = False
1924
result.append((None, base, size))
1929
"""IFF this data access has its own storage area, initialise it.
1933
self._transport.put_bytes_non_atomic(self._filename, '',
1934
mode=self._file_mode)
1936
def open_file(self):
1937
"""IFF this data access can be represented as a single file, open it.
1939
For knits that are not mapped to a single file on disk this will
1942
:return: None or a file handle.
1945
return self._transport.get(self._filename)
1950
def get_raw_records(self, memos_for_retrieval):
1951
"""Get the raw bytes for a records.
1953
:param memos_for_retrieval: An iterable containing the (index, pos,
1954
length) memo for retrieving the bytes. The .knit method ignores
1955
the index as there is always only a single file.
1956
:return: An iterator over the bytes of the records.
1958
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1959
for pos, data in self._transport.readv(self._filename, read_vector):
1963
class _PackAccess(object):
1964
"""Access to knit records via a collection of packs."""
1966
def __init__(self, index_to_packs, writer=None):
1967
"""Create a _PackAccess object.
1969
:param index_to_packs: A dict mapping index objects to the transport
1970
and file names for obtaining data.
1971
:param writer: A tuple (pack.ContainerWriter, write_index) which
1972
contains the pack to write, and the index that reads from it will
1976
self.container_writer = writer[0]
1977
self.write_index = writer[1]
1979
self.container_writer = None
1980
self.write_index = None
1981
self.indices = index_to_packs
1983
def add_raw_records(self, sizes, raw_data):
1984
"""Add raw knit bytes to a storage area.
1986
The data is spooled to the container writer in one bytes-record per
1989
:param sizes: An iterable containing the size of each raw data segment.
1990
:param raw_data: A bytestring containing the data.
1991
:return: A list of memos to retrieve the record later. Each memo is a
1992
tuple - (index, pos, length), where the index field is the
1993
write_index object supplied to the PackAccess object.
1995
assert type(raw_data) == str, \
1996
'data must be plain bytes was %s' % type(raw_data)
2000
p_offset, p_length = self.container_writer.add_bytes_record(
2001
raw_data[offset:offset+size], [])
2003
result.append((self.write_index, p_offset, p_length))
2007
"""Pack based knits do not get individually created."""
2009
def get_raw_records(self, memos_for_retrieval):
2010
"""Get the raw bytes for a records.
2012
:param memos_for_retrieval: An iterable containing the (index, pos,
2013
length) memo for retrieving the bytes. The Pack access method
2014
looks up the pack to use for a given record in its index_to_pack
2016
:return: An iterator over the bytes of the records.
2018
# first pass, group into same-index requests
2020
current_index = None
2021
for (index, offset, length) in memos_for_retrieval:
2022
if current_index == index:
2023
current_list.append((offset, length))
2025
if current_index is not None:
2026
request_lists.append((current_index, current_list))
2027
current_index = index
2028
current_list = [(offset, length)]
2029
# handle the last entry
2030
if current_index is not None:
2031
request_lists.append((current_index, current_list))
2032
for index, offsets in request_lists:
2033
transport, path = self.indices[index]
2034
reader = pack.make_readv_reader(transport, path, offsets)
2035
for names, read_func in reader.iter_records():
2036
yield read_func(None)
2038
def open_file(self):
2039
"""Pack based knits have no single file."""
2042
def set_writer(self, writer, index, (transport, packname)):
2043
"""Set a writer to use for adding data."""
2044
if index is not None:
2045
self.indices[index] = (transport, packname)
2046
self.container_writer = writer
2047
self.write_index = index
2050
class _StreamAccess(object):
2051
"""A Knit Access object that provides data from a datastream.
2053
It also provides a fallback to present as unannotated data, annotated data
2054
from a *backing* access object.
2056
This is triggered by a index_memo which is pointing to a different index
2057
than this was constructed with, and is used to allow extracting full
2058
unannotated texts for insertion into annotated knits.
2061
def __init__(self, reader_callable, stream_index, backing_knit,
2063
"""Create a _StreamAccess object.
2065
:param reader_callable: The reader_callable from the datastream.
2066
This is called to buffer all the data immediately, for
2068
:param stream_index: The index the data stream this provides access to
2069
which will be present in native index_memo's.
2070
:param backing_knit: The knit object that will provide access to
2071
annotated texts which are not available in the stream, so as to
2072
create unannotated texts.
2073
:param orig_factory: The original content factory used to generate the
2074
stream. This is used for checking whether the thunk code for
2075
supporting _copy_texts will generate the correct form of data.
2077
self.data = reader_callable(None)
2078
self.stream_index = stream_index
2079
self.backing_knit = backing_knit
2080
self.orig_factory = orig_factory
2082
def get_raw_records(self, memos_for_retrieval):
2083
"""Get the raw bytes for a records.
2085
:param memos_for_retrieval: An iterable containing the (thunk_flag,
2086
index, start, end) memo for retrieving the bytes.
2087
:return: An iterator over the bytes of the records.
2089
# use a generator for memory friendliness
2090
for thunk_flag, version_id, start, end in memos_for_retrieval:
2091
if version_id is self.stream_index:
2092
yield self.data[start:end]
2094
# we have been asked to thunk. This thunking only occurs when
2095
# we are obtaining plain texts from an annotated backing knit
2096
# so that _copy_texts will work.
2097
# We could improve performance here by scanning for where we need
2098
# to do this and using get_line_list, then interleaving the output
2099
# as desired. However, for now, this is sufficient.
2100
if self.orig_factory.__class__ != KnitPlainFactory:
2101
raise errors.KnitCorrupt(
2102
self, 'Bad thunk request %r' % version_id)
2103
lines = self.backing_knit.get_lines(version_id)
2104
line_bytes = ''.join(lines)
2105
digest = sha_string(line_bytes)
2107
if lines[-1][-1] != '\n':
2108
lines[-1] = lines[-1] + '\n'
2110
orig_options = list(self.backing_knit._index.get_options(version_id))
2111
if 'fulltext' not in orig_options:
2112
if 'line-delta' not in orig_options:
2113
raise errors.KnitCorrupt(self,
2114
'Unknown compression method %r' % orig_options)
2115
orig_options.remove('line-delta')
2116
orig_options.append('fulltext')
2117
# We want plain data, because we expect to thunk only to allow text
2119
size, bytes = self.backing_knit._data._record_to_data(version_id,
2120
digest, lines, line_bytes)
2124
class _StreamIndex(object):
2125
"""A Knit Index object that uses the data map from a datastream."""
2127
def __init__(self, data_list):
2128
"""Create a _StreamIndex object.
2130
:param data_list: The data_list from the datastream.
2132
self.data_list = data_list
2133
self._by_version = {}
2135
for key, options, length, parents in data_list:
2136
self._by_version[key] = options, (pos, pos + length), parents
2139
def get_ancestry(self, versions, topo_sorted):
2140
"""Get an ancestry list for versions."""
2142
# Not needed for basic joins
2143
raise NotImplementedError(self.get_ancestry)
2144
# get a graph of all the mentioned versions:
2145
# Little ugly - basically copied from KnitIndex, but don't want to
2146
# accidentally incorporate too much of that index's code.
2148
pending = set(versions)
2149
cache = self._by_version
2151
version = pending.pop()
2154
parents = [p for p in cache[version][2] if p in cache]
2156
raise RevisionNotPresent(version, self)
2157
# if not completed and not a ghost
2158
pending.update([p for p in parents if p not in ancestry])
2159
ancestry.add(version)
2160
return list(ancestry)
2162
def get_method(self, version_id):
2163
"""Return compression method of specified version."""
2165
options = self._by_version[version_id][0]
2167
# Strictly speaking this should check in the backing knit, but
2168
# until we have a test to discriminate, this will do.
2170
if 'fulltext' in options:
2172
elif 'line-delta' in options:
2175
raise errors.KnitIndexUnknownMethod(self, options)
2177
def get_options(self, version_id):
2178
"""Return a list representing options.
2182
return self._by_version[version_id][0]
2184
def get_parents_with_ghosts(self, version_id):
2185
"""Return parents of specified version with ghosts."""
2186
return self._by_version[version_id][2]
2188
def get_position(self, version_id):
2189
"""Return details needed to access the version.
2191
_StreamAccess has the data as a big array, so we return slice
2192
coordinates into that (as index_memo's are opaque outside the
2193
index and matching access class).
2195
:return: a tuple (thunk_flag, index, start, end). If thunk_flag is
2196
False, index will be self, otherwise it will be a version id.
2199
start, end = self._by_version[version_id][1]
2200
return False, self, start, end
2202
# Signal to the access object to handle this from the backing knit.
2203
return (True, version_id, None, None)
2205
def get_versions(self):
2206
"""Get all the versions in the stream."""
2207
return self._by_version.keys()
2209
def iter_parents(self, version_ids):
2210
"""Iterate through the parents for many version ids.
2212
:param version_ids: An iterable yielding version_ids.
2213
:return: An iterator that yields (version_id, parents). Requested
2214
version_ids not present in the versioned file are simply skipped.
2215
The order is undefined, allowing for different optimisations in
2216
the underlying implementation.
2219
for version in version_ids:
2221
result.append((version, self._by_version[version][2]))
2227
class _KnitData(object):
2228
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2230
The KnitData class provides the logic for parsing and using knit records,
2231
making use of an access method for the low level read and write operations.
2234
def __init__(self, access):
2235
"""Create a KnitData object.
2237
:param access: The access method to use. Access methods such as
2238
_KnitAccess manage the insertion of raw records and the subsequent
2239
retrieval of the same.
2241
self._access = access
2242
self._checked = False
2243
# TODO: jam 20060713 conceptually, this could spill to disk
2244
# if the cached size gets larger than a certain amount
2245
# but it complicates the model a bit, so for now just use
2246
# a simple dictionary
2248
self._do_cache = False
2250
def enable_cache(self):
2251
"""Enable caching of reads."""
2252
self._do_cache = True
2254
def clear_cache(self):
2255
"""Clear the record cache."""
2256
self._do_cache = False
2259
def _open_file(self):
2260
return self._access.open_file()
2262
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2263
"""Convert version_id, digest, lines into a raw data block.
2265
:param dense_lines: The bytes of lines but in a denser form. For
2266
instance, if lines is a list of 1000 bytestrings each ending in \n,
2267
dense_lines may be a list with one line in it, containing all the
2268
1000's lines and their \n's. Using dense_lines if it is already
2269
known is a win because the string join to create bytes in this
2270
function spends less time resizing the final string.
2271
:return: (len, a StringIO instance with the raw data ready to read.)
2273
# Note: using a string copy here increases memory pressure with e.g.
2274
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2275
# when doing the initial commit of a mozilla tree. RBC 20070921
2276
bytes = ''.join(chain(
2277
["version %s %d %s\n" % (version_id,
2280
dense_lines or lines,
2281
["end %s\n" % version_id]))
2282
assert bytes.__class__ == str
2283
compressed_bytes = bytes_to_gzip(bytes)
2284
return len(compressed_bytes), compressed_bytes
2286
def add_raw_records(self, sizes, raw_data):
2287
"""Append a prepared record to the data file.
2289
:param sizes: An iterable containing the size of each raw data segment.
2290
:param raw_data: A bytestring containing the data.
2291
:return: a list of index data for the way the data was stored.
2292
See the access method add_raw_records documentation for more
2295
return self._access.add_raw_records(sizes, raw_data)
2297
def _parse_record_header(self, version_id, raw_data):
2298
"""Parse a record header for consistency.
2300
:return: the header and the decompressor stream.
2301
as (stream, header_record)
2303
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2305
rec = self._check_header(version_id, df.readline())
2306
except Exception, e:
2307
raise KnitCorrupt(self._access,
2308
"While reading {%s} got %s(%s)"
2309
% (version_id, e.__class__.__name__, str(e)))
2312
def _check_header(self, version_id, line):
2315
raise KnitCorrupt(self._access,
2316
'unexpected number of elements in record header')
2317
if rec[1] != version_id:
2318
raise KnitCorrupt(self._access,
2319
'unexpected version, wanted %r, got %r'
2320
% (version_id, rec[1]))
2323
def _parse_record(self, version_id, data):
2325
# 4168 calls in 2880 217 internal
2326
# 4168 calls to _parse_record_header in 2121
2327
# 4168 calls to readlines in 330
2328
df = GzipFile(mode='rb', fileobj=StringIO(data))
2331
record_contents = df.readlines()
2332
except Exception, e:
2333
raise KnitCorrupt(self._access,
2334
"While reading {%s} got %s(%s)"
2335
% (version_id, e.__class__.__name__, str(e)))
2336
header = record_contents.pop(0)
2337
rec = self._check_header(version_id, header)
2339
last_line = record_contents.pop()
2340
if len(record_contents) != int(rec[2]):
2341
raise KnitCorrupt(self._access,
2342
'incorrect number of lines %s != %s'
2344
% (len(record_contents), int(rec[2]),
2346
if last_line != 'end %s\n' % rec[1]:
2347
raise KnitCorrupt(self._access,
2348
'unexpected version end line %r, wanted %r'
2349
% (last_line, version_id))
2351
return record_contents, rec[3]
2353
def read_records_iter_raw(self, records):
2354
"""Read text records from data file and yield raw data.
2356
This unpacks enough of the text record to validate the id is
2357
as expected but thats all.
2359
# setup an iterator of the external records:
2360
# uses readv so nice and fast we hope.
2362
# grab the disk data needed.
2364
# Don't check _cache if it is empty
2365
needed_offsets = [index_memo for version_id, index_memo
2367
if version_id not in self._cache]
2369
needed_offsets = [index_memo for version_id, index_memo
2372
raw_records = self._access.get_raw_records(needed_offsets)
2374
for version_id, index_memo in records:
2375
if version_id in self._cache:
2376
# This data has already been validated
2377
data = self._cache[version_id]
2379
data = raw_records.next()
2381
self._cache[version_id] = data
2383
# validate the header
2384
df, rec = self._parse_record_header(version_id, data)
2386
yield version_id, data
2388
def read_records_iter(self, records):
2389
"""Read text records from data file and yield result.
2391
The result will be returned in whatever is the fastest to read.
2392
Not by the order requested. Also, multiple requests for the same
2393
record will only yield 1 response.
2394
:param records: A list of (version_id, pos, len) entries
2395
:return: Yields (version_id, contents, digest) in the order
2396
read, not the order requested
2402
# Skip records we have alread seen
2403
yielded_records = set()
2404
needed_records = set()
2405
for record in records:
2406
if record[0] in self._cache:
2407
if record[0] in yielded_records:
2409
yielded_records.add(record[0])
2410
data = self._cache[record[0]]
2411
content, digest = self._parse_record(record[0], data)
2412
yield (record[0], content, digest)
2414
needed_records.add(record)
2415
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2417
needed_records = sorted(set(records), key=operator.itemgetter(1))
2419
if not needed_records:
2422
# The transport optimizes the fetching as well
2423
# (ie, reads continuous ranges.)
2424
raw_data = self._access.get_raw_records(
2425
[index_memo for version_id, index_memo in needed_records])
2427
for (version_id, index_memo), data in \
2428
izip(iter(needed_records), raw_data):
2429
content, digest = self._parse_record(version_id, data)
2431
self._cache[version_id] = data
2432
yield version_id, content, digest
2434
def read_records(self, records):
2435
"""Read records into a dictionary."""
2437
for record_id, content, digest in \
2438
self.read_records_iter(records):
2439
components[record_id] = (content, digest)
2443
class InterKnit(InterVersionedFile):
2444
"""Optimised code paths for knit to knit operations."""
2446
_matching_file_from_factory = KnitVersionedFile
2447
_matching_file_to_factory = KnitVersionedFile
2450
def is_compatible(source, target):
2451
"""Be compatible with knits. """
2453
return (isinstance(source, KnitVersionedFile) and
2454
isinstance(target, KnitVersionedFile))
2455
except AttributeError:
2458
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2459
"""Copy texts to the target by extracting and adding them one by one.
2461
see join() for the parameter definitions.
2463
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2464
graph = self.source.get_graph(version_ids)
2465
order = topo_sort(graph.items())
2467
def size_of_content(content):
2468
return sum(len(line) for line in content.text())
2469
# Cache at most 10MB of parent texts
2470
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2471
compute_size=size_of_content)
2472
# TODO: jam 20071116 It would be nice to have a streaming interface to
2473
# get multiple texts from a source. The source could be smarter
2474
# about how it handled intermediate stages.
2475
# get_line_list() or make_mpdiffs() seem like a possibility, but
2476
# at the moment they extract all full texts into memory, which
2477
# causes us to store more than our 3x fulltext goal.
2478
# Repository.iter_files_bytes() may be another possibility
2479
to_process = [version for version in order
2480
if version not in self.target]
2481
total = len(to_process)
2482
pb = ui.ui_factory.nested_progress_bar()
2484
for index, version in enumerate(to_process):
2485
pb.update('Converting versioned data', index, total)
2486
sha1, num_bytes, parent_text = self.target.add_lines(version,
2487
self.source.get_parents_with_ghosts(version),
2488
self.source.get_lines(version),
2489
parent_texts=parent_cache)
2490
parent_cache[version] = parent_text
2495
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2496
"""See InterVersionedFile.join."""
2497
assert isinstance(self.source, KnitVersionedFile)
2498
assert isinstance(self.target, KnitVersionedFile)
2500
# If the source and target are mismatched w.r.t. annotations vs
2501
# plain, the data needs to be converted accordingly
2502
if self.source.factory.annotated == self.target.factory.annotated:
2504
elif self.source.factory.annotated:
2505
converter = self._anno_to_plain_converter
2507
# We're converting from a plain to an annotated knit. Copy them
2508
# across by full texts.
2509
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2511
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2515
pb = ui.ui_factory.nested_progress_bar()
2517
version_ids = list(version_ids)
2518
if None in version_ids:
2519
version_ids.remove(None)
2521
self.source_ancestry = set(self.source.get_ancestry(version_ids,
2523
this_versions = set(self.target._index.get_versions())
2524
# XXX: For efficiency we should not look at the whole index,
2525
# we only need to consider the referenced revisions - they
2526
# must all be present, or the method must be full-text.
2527
# TODO, RBC 20070919
2528
needed_versions = self.source_ancestry - this_versions
2530
if not needed_versions:
2532
full_list = topo_sort(self.source.get_graph())
2534
version_list = [i for i in full_list if (not self.target.has_version(i)
2535
and i in needed_versions)]
2539
copy_queue_records = []
2541
for version_id in version_list:
2542
options = self.source._index.get_options(version_id)
2543
parents = self.source._index.get_parents_with_ghosts(version_id)
2544
# check that its will be a consistent copy:
2545
for parent in parents:
2546
# if source has the parent, we must :
2547
# * already have it or
2548
# * have it scheduled already
2549
# otherwise we don't care
2550
assert (self.target.has_version(parent) or
2551
parent in copy_set or
2552
not self.source.has_version(parent))
2553
index_memo = self.source._index.get_position(version_id)
2554
copy_queue_records.append((version_id, index_memo))
2555
copy_queue.append((version_id, options, parents))
2556
copy_set.add(version_id)
2558
# data suck the join:
2560
total = len(version_list)
2563
for (version_id, raw_data), \
2564
(version_id2, options, parents) in \
2565
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2567
assert version_id == version_id2, 'logic error, inconsistent results'
2569
pb.update("Joining knit", count, total)
2571
size, raw_data = converter(raw_data, version_id, options,
2574
size = len(raw_data)
2575
raw_records.append((version_id, options, parents, size))
2576
raw_datum.append(raw_data)
2577
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2582
def _anno_to_plain_converter(self, raw_data, version_id, options,
2584
"""Convert annotated content to plain content."""
2585
data, digest = self.source._data._parse_record(version_id, raw_data)
2586
if 'fulltext' in options:
2587
content = self.source.factory.parse_fulltext(data, version_id)
2588
lines = self.target.factory.lower_fulltext(content)
2590
delta = self.source.factory.parse_line_delta(data, version_id,
2592
lines = self.target.factory.lower_line_delta(delta)
2593
return self.target._data._record_to_data(version_id, digest, lines)
2596
InterVersionedFile.register_optimiser(InterKnit)
2599
class WeaveToKnit(InterVersionedFile):
2600
"""Optimised code paths for weave to knit operations."""
2602
_matching_file_from_factory = bzrlib.weave.WeaveFile
2603
_matching_file_to_factory = KnitVersionedFile
2606
def is_compatible(source, target):
2607
"""Be compatible with weaves to knits."""
2609
return (isinstance(source, bzrlib.weave.Weave) and
2610
isinstance(target, KnitVersionedFile))
2611
except AttributeError:
2614
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2615
"""See InterVersionedFile.join."""
2616
assert isinstance(self.source, bzrlib.weave.Weave)
2617
assert isinstance(self.target, KnitVersionedFile)
2619
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2624
pb = ui.ui_factory.nested_progress_bar()
2626
version_ids = list(version_ids)
2628
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2629
this_versions = set(self.target._index.get_versions())
2630
needed_versions = self.source_ancestry - this_versions
2632
if not needed_versions:
2634
full_list = topo_sort(self.source.get_graph())
2636
version_list = [i for i in full_list if (not self.target.has_version(i)
2637
and i in needed_versions)]
2641
total = len(version_list)
2642
for version_id in version_list:
2643
pb.update("Converting to knit", count, total)
2644
parents = self.source.get_parents(version_id)
2645
# check that its will be a consistent copy:
2646
for parent in parents:
2647
# if source has the parent, we must already have it
2648
assert (self.target.has_version(parent))
2649
self.target.add_lines(
2650
version_id, parents, self.source.get_lines(version_id))
2657
InterVersionedFile.register_optimiser(WeaveToKnit)
2660
# Deprecated, use PatienceSequenceMatcher instead
2661
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2664
def annotate_knit(knit, revision_id):
2665
"""Annotate a knit with no cached annotations.
2667
This implementation is for knits with no cached annotations.
2668
It will work for knits with cached annotations, but this is not
2671
ancestry = knit.get_ancestry(revision_id)
2672
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2674
for candidate in ancestry:
2675
if candidate in annotations:
2677
parents = knit.get_parents(candidate)
2678
if len(parents) == 0:
2680
elif knit._index.get_method(candidate) != 'line-delta':
2683
parent, sha1, noeol, delta = knit.get_delta(candidate)
2684
blocks = KnitContent.get_line_delta_blocks(delta,
2685
fulltext[parents[0]], fulltext[candidate])
2686
annotations[candidate] = list(annotate.reannotate([annotations[p]
2687
for p in parents], fulltext[candidate], candidate, blocks))
2688
return iter(annotations[revision_id])
2692
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2694
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
