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, index_memo, 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
pending_components = version_ids
886
while pending_components:
887
build_details = self._index.get_build_details(pending_components)
888
pending_components = set()
889
for version_id, details in build_details.items():
890
method, index_memo, compression_parent = details
891
if compression_parent is not None:
892
pending_components.add(compression_parent)
893
component_data[version_id] = details
894
return component_data
896
def _get_content(self, version_id, parent_texts={}):
897
"""Returns a content object that makes up the specified
899
cached_version = parent_texts.get(version_id, None)
900
if cached_version is not None:
901
if not self.has_version(version_id):
902
raise RevisionNotPresent(version_id, self.filename)
903
return cached_version
905
text_map, contents_map = self._get_content_maps([version_id])
906
return contents_map[version_id]
908
def _check_versions_present(self, version_ids):
909
"""Check that all specified versions are present."""
910
self._index.check_versions_present(version_ids)
912
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
913
nostore_sha, random_id, check_content):
914
"""See VersionedFile.add_lines_with_ghosts()."""
915
self._check_add(version_id, lines, random_id, check_content)
916
return self._add(version_id, lines, parents, self.delta,
917
parent_texts, None, nostore_sha, random_id)
919
def _add_lines(self, version_id, parents, lines, parent_texts,
920
left_matching_blocks, nostore_sha, random_id, check_content):
921
"""See VersionedFile.add_lines."""
922
self._check_add(version_id, lines, random_id, check_content)
923
self._check_versions_present(parents)
924
return self._add(version_id, lines[:], parents, self.delta,
925
parent_texts, left_matching_blocks, nostore_sha, random_id)
927
def _check_add(self, version_id, lines, random_id, check_content):
928
"""check that version_id and lines are safe to add."""
929
if contains_whitespace(version_id):
930
raise InvalidRevisionId(version_id, self.filename)
931
self.check_not_reserved_id(version_id)
932
# Technically this could be avoided if we are happy to allow duplicate
933
# id insertion when other things than bzr core insert texts, but it
934
# seems useful for folk using the knit api directly to have some safety
935
# blanket that we can disable.
936
if not random_id and self.has_version(version_id):
937
raise RevisionAlreadyPresent(version_id, self.filename)
939
self._check_lines_not_unicode(lines)
940
self._check_lines_are_lines(lines)
942
def _add(self, version_id, lines, parents, delta, parent_texts,
943
left_matching_blocks, nostore_sha, random_id):
944
"""Add a set of lines on top of version specified by parents.
946
If delta is true, compress the text as a line-delta against
949
Any versions not present will be converted into ghosts.
951
# first thing, if the content is something we don't need to store, find
953
line_bytes = ''.join(lines)
954
digest = sha_string(line_bytes)
955
if nostore_sha == digest:
956
raise errors.ExistingContent
959
if parent_texts is None:
961
for parent in parents:
962
if self.has_version(parent):
963
present_parents.append(parent)
965
# can only compress against the left most present parent.
967
(len(present_parents) == 0 or
968
present_parents[0] != parents[0])):
971
text_length = len(line_bytes)
974
if lines[-1][-1] != '\n':
975
# copy the contents of lines.
977
options.append('no-eol')
978
lines[-1] = lines[-1] + '\n'
982
# To speed the extract of texts the delta chain is limited
983
# to a fixed number of deltas. This should minimize both
984
# I/O and the time spend applying deltas.
985
delta = self._check_should_delta(present_parents)
987
assert isinstance(version_id, str)
988
content = self.factory.make(lines, version_id)
989
if delta or (self.factory.annotated and len(present_parents) > 0):
990
# Merge annotations from parent texts if needed.
991
delta_hunks = self._merge_annotations(content, present_parents,
992
parent_texts, delta, self.factory.annotated,
993
left_matching_blocks)
996
options.append('line-delta')
997
store_lines = self.factory.lower_line_delta(delta_hunks)
998
size, bytes = self._data._record_to_data(version_id, digest,
1001
options.append('fulltext')
1002
# isinstance is slower and we have no hierarchy.
1003
if self.factory.__class__ == KnitPlainFactory:
1004
# Use the already joined bytes saving iteration time in
1006
size, bytes = self._data._record_to_data(version_id, digest,
1007
lines, [line_bytes])
1009
# get mixed annotation + content and feed it into the
1011
store_lines = self.factory.lower_fulltext(content)
1012
size, bytes = self._data._record_to_data(version_id, digest,
1015
access_memo = self._data.add_raw_records([size], bytes)[0]
1016
self._index.add_versions(
1017
((version_id, options, access_memo, parents),),
1018
random_id=random_id)
1019
return digest, text_length, content
1021
def check(self, progress_bar=None):
1022
"""See VersionedFile.check()."""
1024
def _clone_text(self, new_version_id, old_version_id, parents):
1025
"""See VersionedFile.clone_text()."""
1026
# FIXME RBC 20060228 make fast by only inserting an index with null
1028
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
1030
def get_lines(self, version_id):
1031
"""See VersionedFile.get_lines()."""
1032
return self.get_line_list([version_id])[0]
1034
def _get_record_map(self, version_ids):
1035
"""Produce a dictionary of knit records.
1037
The keys are version_ids, the values are tuples of (method, content,
1039
method is the way the content should be applied.
1040
content is a KnitContent object.
1041
digest is the SHA1 digest of this version id after all steps are done
1042
next is the build-parent of the version, i.e. the leftmost ancestor.
1043
If the method is fulltext, next will be None.
1045
position_map = self._get_components_positions(version_ids)
1046
# c = component_id, m = method, i_m = index_memo, n = next
1047
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
1049
for component_id, content, digest in \
1050
self._data.read_records_iter(records):
1051
method, index_memo, next = position_map[component_id]
1052
record_map[component_id] = method, content, digest, next
1056
def get_text(self, version_id):
1057
"""See VersionedFile.get_text"""
1058
return self.get_texts([version_id])[0]
1060
def get_texts(self, version_ids):
1061
return [''.join(l) for l in self.get_line_list(version_ids)]
1063
def get_line_list(self, version_ids):
1064
"""Return the texts of listed versions as a list of strings."""
1065
for version_id in version_ids:
1066
self.check_not_reserved_id(version_id)
1067
text_map, content_map = self._get_content_maps(version_ids)
1068
return [text_map[v] for v in version_ids]
1070
_get_lf_split_line_list = get_line_list
1072
def _get_content_maps(self, version_ids):
1073
"""Produce maps of text and KnitContents
1075
:return: (text_map, content_map) where text_map contains the texts for
1076
the requested versions and content_map contains the KnitContents.
1077
Both dicts take version_ids as their keys.
1079
# FUTURE: This function could be improved for the 'extract many' case
1080
# by tracking each component and only doing the copy when the number of
1081
# children than need to apply delta's to it is > 1 or it is part of the
1083
version_ids = list(version_ids)
1084
multiple_versions = len(version_ids) != 1
1085
record_map = self._get_record_map(version_ids)
1090
for version_id in version_ids:
1093
while cursor is not None:
1094
method, data, digest, next = record_map[cursor]
1095
components.append((cursor, method, data, digest))
1096
if cursor in content_map:
1101
for component_id, method, data, digest in reversed(components):
1102
if component_id in content_map:
1103
content = content_map[component_id]
1105
if method == 'fulltext':
1106
assert content is None
1107
content = self.factory.parse_fulltext(data, version_id)
1108
elif method == 'line-delta':
1109
delta = self.factory.parse_line_delta(data, version_id)
1110
if multiple_versions:
1111
# only doing this when we want multiple versions
1112
# output avoids list copies - which reference and
1113
# dereference many strings.
1114
content = content.copy()
1115
content.apply_delta(delta, version_id)
1116
if multiple_versions:
1117
content_map[component_id] = content
1119
if 'no-eol' in self._index.get_options(version_id):
1120
if multiple_versions:
1121
content = content.copy()
1122
content.strip_last_line_newline()
1123
final_content[version_id] = content
1125
# digest here is the digest from the last applied component.
1126
text = content.text()
1127
actual_sha = sha_strings(text)
1128
if actual_sha != digest:
1129
raise KnitCorrupt(self.filename,
1131
'\n of reconstructed text does not match'
1133
'\n for version %s' %
1134
(actual_sha, digest, version_id))
1135
text_map[version_id] = text
1136
return text_map, final_content
1138
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1140
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1141
if version_ids is None:
1142
version_ids = self.versions()
1144
pb = progress.DummyProgress()
1145
# we don't care about inclusions, the caller cares.
1146
# but we need to setup a list of records to visit.
1147
# we need version_id, position, length
1148
version_id_records = []
1149
requested_versions = set(version_ids)
1150
# filter for available versions
1151
for version_id in requested_versions:
1152
if not self.has_version(version_id):
1153
raise RevisionNotPresent(version_id, self.filename)
1154
# get a in-component-order queue:
1155
for version_id in self.versions():
1156
if version_id in requested_versions:
1157
index_memo = self._index.get_position(version_id)
1158
version_id_records.append((version_id, index_memo))
1160
total = len(version_id_records)
1161
for version_idx, (version_id, data, sha_value) in \
1162
enumerate(self._data.read_records_iter(version_id_records)):
1163
pb.update('Walking content.', version_idx, total)
1164
method = self._index.get_method(version_id)
1166
assert method in ('fulltext', 'line-delta')
1167
if method == 'fulltext':
1168
line_iterator = self.factory.get_fulltext_content(data)
1170
line_iterator = self.factory.get_linedelta_content(data)
1171
# XXX: It might be more efficient to yield (version_id,
1172
# line_iterator) in the future. However for now, this is a simpler
1173
# change to integrate into the rest of the codebase. RBC 20071110
1174
for line in line_iterator:
1175
yield line, version_id
1177
pb.update('Walking content.', total, total)
1179
def iter_parents(self, version_ids):
1180
"""Iterate through the parents for many version ids.
1182
:param version_ids: An iterable yielding version_ids.
1183
:return: An iterator that yields (version_id, parents). Requested
1184
version_ids not present in the versioned file are simply skipped.
1185
The order is undefined, allowing for different optimisations in
1186
the underlying implementation.
1188
return self._index.iter_parents(version_ids)
1190
def num_versions(self):
1191
"""See VersionedFile.num_versions()."""
1192
return self._index.num_versions()
1194
__len__ = num_versions
1196
def annotate_iter(self, version_id):
1197
"""See VersionedFile.annotate_iter."""
1198
return self.factory.annotate_iter(self, version_id)
1200
def get_parents(self, version_id):
1201
"""See VersionedFile.get_parents."""
1204
# 52554 calls in 1264 872 internal down from 3674
1206
return self._index.get_parents(version_id)
1208
raise RevisionNotPresent(version_id, self.filename)
1210
def get_parents_with_ghosts(self, version_id):
1211
"""See VersionedFile.get_parents."""
1213
return self._index.get_parents_with_ghosts(version_id)
1215
raise RevisionNotPresent(version_id, self.filename)
1217
def get_ancestry(self, versions, topo_sorted=True):
1218
"""See VersionedFile.get_ancestry."""
1219
if isinstance(versions, basestring):
1220
versions = [versions]
1223
return self._index.get_ancestry(versions, topo_sorted)
1225
def get_ancestry_with_ghosts(self, versions):
1226
"""See VersionedFile.get_ancestry_with_ghosts."""
1227
if isinstance(versions, basestring):
1228
versions = [versions]
1231
return self._index.get_ancestry_with_ghosts(versions)
1233
def plan_merge(self, ver_a, ver_b):
1234
"""See VersionedFile.plan_merge."""
1235
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1236
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1237
annotated_a = self.annotate(ver_a)
1238
annotated_b = self.annotate(ver_b)
1239
return merge._plan_annotate_merge(annotated_a, annotated_b,
1240
ancestors_a, ancestors_b)
1243
class _KnitComponentFile(object):
1244
"""One of the files used to implement a knit database"""
1246
def __init__(self, transport, filename, mode, file_mode=None,
1247
create_parent_dir=False, dir_mode=None):
1248
self._transport = transport
1249
self._filename = filename
1251
self._file_mode = file_mode
1252
self._dir_mode = dir_mode
1253
self._create_parent_dir = create_parent_dir
1254
self._need_to_create = False
1256
def _full_path(self):
1257
"""Return the full path to this file."""
1258
return self._transport.base + self._filename
1260
def check_header(self, fp):
1261
line = fp.readline()
1263
# An empty file can actually be treated as though the file doesn't
1265
raise errors.NoSuchFile(self._full_path())
1266
if line != self.HEADER:
1267
raise KnitHeaderError(badline=line,
1268
filename=self._transport.abspath(self._filename))
1271
return '%s(%s)' % (self.__class__.__name__, self._filename)
1274
class _KnitIndex(_KnitComponentFile):
1275
"""Manages knit index file.
1277
The index is already kept in memory and read on startup, to enable
1278
fast lookups of revision information. The cursor of the index
1279
file is always pointing to the end, making it easy to append
1282
_cache is a cache for fast mapping from version id to a Index
1285
_history is a cache for fast mapping from indexes to version ids.
1287
The index data format is dictionary compressed when it comes to
1288
parent references; a index entry may only have parents that with a
1289
lover index number. As a result, the index is topological sorted.
1291
Duplicate entries may be written to the index for a single version id
1292
if this is done then the latter one completely replaces the former:
1293
this allows updates to correct version and parent information.
1294
Note that the two entries may share the delta, and that successive
1295
annotations and references MUST point to the first entry.
1297
The index file on disc contains a header, followed by one line per knit
1298
record. The same revision can be present in an index file more than once.
1299
The first occurrence gets assigned a sequence number starting from 0.
1301
The format of a single line is
1302
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1303
REVISION_ID is a utf8-encoded revision id
1304
FLAGS is a comma separated list of flags about the record. Values include
1305
no-eol, line-delta, fulltext.
1306
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1307
that the the compressed data starts at.
1308
LENGTH is the ascii representation of the length of the data file.
1309
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1311
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1312
revision id already in the knit that is a parent of REVISION_ID.
1313
The ' :' marker is the end of record marker.
1316
when a write is interrupted to the index file, it will result in a line
1317
that does not end in ' :'. If the ' :' is not present at the end of a line,
1318
or at the end of the file, then the record that is missing it will be
1319
ignored by the parser.
1321
When writing new records to the index file, the data is preceded by '\n'
1322
to ensure that records always start on new lines even if the last write was
1323
interrupted. As a result its normal for the last line in the index to be
1324
missing a trailing newline. One can be added with no harmful effects.
1327
HEADER = "# bzr knit index 8\n"
1329
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1330
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1332
def _cache_version(self, version_id, options, pos, size, parents):
1333
"""Cache a version record in the history array and index cache.
1335
This is inlined into _load_data for performance. KEEP IN SYNC.
1336
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1339
# only want the _history index to reference the 1st index entry
1341
if version_id not in self._cache:
1342
index = len(self._history)
1343
self._history.append(version_id)
1345
index = self._cache[version_id][5]
1346
self._cache[version_id] = (version_id,
1353
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1354
create_parent_dir=False, delay_create=False, dir_mode=None):
1355
_KnitComponentFile.__init__(self, transport, filename, mode,
1356
file_mode=file_mode,
1357
create_parent_dir=create_parent_dir,
1360
# position in _history is the 'official' index for a revision
1361
# but the values may have come from a newer entry.
1362
# so - wc -l of a knit index is != the number of unique names
1366
fp = self._transport.get(self._filename)
1368
# _load_data may raise NoSuchFile if the target knit is
1370
_load_data(self, fp)
1374
if mode != 'w' or not create:
1377
self._need_to_create = True
1379
self._transport.put_bytes_non_atomic(
1380
self._filename, self.HEADER, mode=self._file_mode)
1382
def get_graph(self):
1383
"""Return a list of the node:parents lists from this knit index."""
1384
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1386
def get_ancestry(self, versions, topo_sorted=True):
1387
"""See VersionedFile.get_ancestry."""
1388
# get a graph of all the mentioned versions:
1390
pending = set(versions)
1393
version = pending.pop()
1396
parents = [p for p in cache[version][4] if p in cache]
1398
raise RevisionNotPresent(version, self._filename)
1399
# if not completed and not a ghost
1400
pending.update([p for p in parents if p not in graph])
1401
graph[version] = parents
1404
return topo_sort(graph.items())
1406
def get_ancestry_with_ghosts(self, versions):
1407
"""See VersionedFile.get_ancestry_with_ghosts."""
1408
# get a graph of all the mentioned versions:
1409
self.check_versions_present(versions)
1412
pending = set(versions)
1414
version = pending.pop()
1416
parents = cache[version][4]
1422
pending.update([p for p in parents if p not in graph])
1423
graph[version] = parents
1424
return topo_sort(graph.items())
1426
def get_build_details(self, version_ids):
1427
"""Get the method, index_memo and compression parent for version_ids.
1429
:param version_ids: An iterable of version_ids.
1430
:return: A dict of version_id:(method, index_memo, compression_parent).
1433
for version_id in version_ids:
1434
method = self.get_method(version_id)
1435
if method == 'fulltext':
1436
compression_parent = None
1438
compression_parent = self.get_parents_with_ghosts(version_id)[0]
1439
index_memo = self.get_position(version_id)
1440
result[version_id] = (method, index_memo, compression_parent)
1443
def iter_parents(self, version_ids):
1444
"""Iterate through the parents for many version ids.
1446
:param version_ids: An iterable yielding version_ids.
1447
:return: An iterator that yields (version_id, parents). Requested
1448
version_ids not present in the versioned file are simply skipped.
1449
The order is undefined, allowing for different optimisations in
1450
the underlying implementation.
1452
for version_id in version_ids:
1454
yield version_id, tuple(self.get_parents(version_id))
1458
def num_versions(self):
1459
return len(self._history)
1461
__len__ = num_versions
1463
def get_versions(self):
1464
"""Get all the versions in the file. not topologically sorted."""
1465
return self._history
1467
def _version_list_to_index(self, versions):
1470
for version in versions:
1471
if version in cache:
1472
# -- inlined lookup() --
1473
result_list.append(str(cache[version][5]))
1474
# -- end lookup () --
1476
result_list.append('.' + version)
1477
return ' '.join(result_list)
1479
def add_version(self, version_id, options, index_memo, parents):
1480
"""Add a version record to the index."""
1481
self.add_versions(((version_id, options, index_memo, parents),))
1483
def add_versions(self, versions, random_id=False):
1484
"""Add multiple versions to the index.
1486
:param versions: a list of tuples:
1487
(version_id, options, pos, size, parents).
1488
:param random_id: If True the ids being added were randomly generated
1489
and no check for existence will be performed.
1492
orig_history = self._history[:]
1493
orig_cache = self._cache.copy()
1496
for version_id, options, (index, pos, size), parents in versions:
1497
line = "\n%s %s %s %s %s :" % (version_id,
1501
self._version_list_to_index(parents))
1502
assert isinstance(line, str), \
1503
'content must be utf-8 encoded: %r' % (line,)
1505
self._cache_version(version_id, options, pos, size, parents)
1506
if not self._need_to_create:
1507
self._transport.append_bytes(self._filename, ''.join(lines))
1510
sio.write(self.HEADER)
1511
sio.writelines(lines)
1513
self._transport.put_file_non_atomic(self._filename, sio,
1514
create_parent_dir=self._create_parent_dir,
1515
mode=self._file_mode,
1516
dir_mode=self._dir_mode)
1517
self._need_to_create = False
1519
# If any problems happen, restore the original values and re-raise
1520
self._history = orig_history
1521
self._cache = orig_cache
1524
def has_version(self, version_id):
1525
"""True if the version is in the index."""
1526
return version_id in self._cache
1528
def get_position(self, version_id):
1529
"""Return details needed to access the version.
1531
.kndx indices do not support split-out data, so return None for the
1534
:return: a tuple (None, data position, size) to hand to the access
1535
logic to get the record.
1537
entry = self._cache[version_id]
1538
return None, entry[2], entry[3]
1540
def get_method(self, version_id):
1541
"""Return compression method of specified version."""
1543
options = self._cache[version_id][1]
1545
raise RevisionNotPresent(version_id, self._filename)
1546
if 'fulltext' in options:
1549
if 'line-delta' not in options:
1550
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1553
def get_options(self, version_id):
1554
"""Return a list representing options.
1558
return self._cache[version_id][1]
1560
def get_parents(self, version_id):
1561
"""Return parents of specified version ignoring ghosts."""
1562
return [parent for parent in self._cache[version_id][4]
1563
if parent in self._cache]
1565
def get_parents_with_ghosts(self, version_id):
1566
"""Return parents of specified version with ghosts."""
1567
return self._cache[version_id][4]
1569
def check_versions_present(self, version_ids):
1570
"""Check that all specified versions are present."""
1572
for version_id in version_ids:
1573
if version_id not in cache:
1574
raise RevisionNotPresent(version_id, self._filename)
1577
class KnitGraphIndex(object):
1578
"""A knit index that builds on GraphIndex."""
1580
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1581
"""Construct a KnitGraphIndex on a graph_index.
1583
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1584
:param deltas: Allow delta-compressed records.
1585
:param add_callback: If not None, allow additions to the index and call
1586
this callback with a list of added GraphIndex nodes:
1587
[(node, value, node_refs), ...]
1588
:param parents: If True, record knits parents, if not do not record
1591
self._graph_index = graph_index
1592
self._deltas = deltas
1593
self._add_callback = add_callback
1594
self._parents = parents
1595
if deltas and not parents:
1596
raise KnitCorrupt(self, "Cannot do delta compression without "
1599
def _get_entries(self, keys, check_present=False):
1600
"""Get the entries for keys.
1602
:param keys: An iterable of index keys, - 1-tuples.
1607
for node in self._graph_index.iter_entries(keys):
1609
found_keys.add(node[1])
1611
# adapt parentless index to the rest of the code.
1612
for node in self._graph_index.iter_entries(keys):
1613
yield node[0], node[1], node[2], ()
1614
found_keys.add(node[1])
1616
missing_keys = keys.difference(found_keys)
1618
raise RevisionNotPresent(missing_keys.pop(), self)
1620
def _present_keys(self, version_ids):
1622
node[1] for node in self._get_entries(version_ids)])
1624
def _parentless_ancestry(self, versions):
1625
"""Honour the get_ancestry API for parentless knit indices."""
1626
wanted_keys = self._version_ids_to_keys(versions)
1627
present_keys = self._present_keys(wanted_keys)
1628
missing = set(wanted_keys).difference(present_keys)
1630
raise RevisionNotPresent(missing.pop(), self)
1631
return list(self._keys_to_version_ids(present_keys))
1633
def get_ancestry(self, versions, topo_sorted=True):
1634
"""See VersionedFile.get_ancestry."""
1635
if not self._parents:
1636
return self._parentless_ancestry(versions)
1637
# XXX: This will do len(history) index calls - perhaps
1638
# it should be altered to be a index core feature?
1639
# get a graph of all the mentioned versions:
1642
versions = self._version_ids_to_keys(versions)
1643
pending = set(versions)
1645
# get all pending nodes
1646
this_iteration = pending
1647
new_nodes = self._get_entries(this_iteration)
1650
for (index, key, value, node_refs) in new_nodes:
1651
# dont ask for ghosties - otherwise
1652
# we we can end up looping with pending
1653
# being entirely ghosted.
1654
graph[key] = [parent for parent in node_refs[0]
1655
if parent not in ghosts]
1657
for parent in graph[key]:
1658
# dont examine known nodes again
1663
ghosts.update(this_iteration.difference(found))
1664
if versions.difference(graph):
1665
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1667
result_keys = topo_sort(graph.items())
1669
result_keys = graph.iterkeys()
1670
return [key[0] for key in result_keys]
1672
def get_ancestry_with_ghosts(self, versions):
1673
"""See VersionedFile.get_ancestry."""
1674
if not self._parents:
1675
return self._parentless_ancestry(versions)
1676
# XXX: This will do len(history) index calls - perhaps
1677
# it should be altered to be a index core feature?
1678
# get a graph of all the mentioned versions:
1680
versions = self._version_ids_to_keys(versions)
1681
pending = set(versions)
1683
# get all pending nodes
1684
this_iteration = pending
1685
new_nodes = self._get_entries(this_iteration)
1687
for (index, key, value, node_refs) in new_nodes:
1688
graph[key] = node_refs[0]
1690
for parent in graph[key]:
1691
# dont examine known nodes again
1695
missing_versions = this_iteration.difference(graph)
1696
missing_needed = versions.intersection(missing_versions)
1698
raise RevisionNotPresent(missing_needed.pop(), self)
1699
for missing_version in missing_versions:
1700
# add a key, no parents
1701
graph[missing_version] = []
1702
pending.discard(missing_version) # don't look for it
1703
result_keys = topo_sort(graph.items())
1704
return [key[0] for key in result_keys]
1706
def get_build_details(self, version_ids):
1707
"""Get the method, index_memo and compression parent for version_ids.
1709
:param version_ids: An iterable of version_ids.
1710
:return: A dict of version_id:(method, index_memo, compression_parent).
1713
entries = self._get_entries(self._version_ids_to_keys(version_ids), True)
1714
for entry in entries:
1715
version_id = self._keys_to_version_ids((entry[1],))[0]
1716
if not self._deltas:
1717
compression_parent = None
1719
compression_parent_key = self._compression_parent(entry)
1720
if compression_parent_key:
1721
compression_parent = self._keys_to_version_ids(
1722
(compression_parent_key,))[0]
1724
compression_parent = None
1725
if compression_parent:
1726
method = 'line-delta'
1729
result[version_id] = (method, self._node_to_position(entry),
1733
def _compression_parent(self, an_entry):
1734
# return the key that an_entry is compressed against, or None
1735
# Grab the second parent list (as deltas implies parents currently)
1736
compression_parents = an_entry[3][1]
1737
if not compression_parents:
1739
assert len(compression_parents) == 1
1740
return compression_parents[0]
1742
def _get_method(self, node):
1743
if not self._deltas:
1745
if self._compression_parent(node):
1750
def get_graph(self):
1751
"""Return a list of the node:parents lists from this knit index."""
1752
if not self._parents:
1753
return [(key, ()) for key in self.get_versions()]
1755
for index, key, value, refs in self._graph_index.iter_all_entries():
1756
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1759
def iter_parents(self, version_ids):
1760
"""Iterate through the parents for many version ids.
1762
:param version_ids: An iterable yielding version_ids.
1763
:return: An iterator that yields (version_id, parents). Requested
1764
version_ids not present in the versioned file are simply skipped.
1765
The order is undefined, allowing for different optimisations in
1766
the underlying implementation.
1769
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1771
present_parents = set()
1772
for node in all_nodes:
1773
all_parents.update(node[3][0])
1774
# any node we are querying must be present
1775
present_parents.add(node[1])
1776
unknown_parents = all_parents.difference(present_parents)
1777
present_parents.update(self._present_keys(unknown_parents))
1778
for node in all_nodes:
1780
for parent in node[3][0]:
1781
if parent in present_parents:
1782
parents.append(parent[0])
1783
yield node[1][0], tuple(parents)
1785
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1786
yield node[1][0], ()
1788
def num_versions(self):
1789
return len(list(self._graph_index.iter_all_entries()))
1791
__len__ = num_versions
1793
def get_versions(self):
1794
"""Get all the versions in the file. not topologically sorted."""
1795
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1797
def has_version(self, version_id):
1798
"""True if the version is in the index."""
1799
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1801
def _keys_to_version_ids(self, keys):
1802
return tuple(key[0] for key in keys)
1804
def get_position(self, version_id):
1805
"""Return details needed to access the version.
1807
:return: a tuple (index, data position, size) to hand to the access
1808
logic to get the record.
1810
node = self._get_node(version_id)
1811
return self._node_to_position(node)
1813
def _node_to_position(self, node):
1814
"""Convert an index value to position details."""
1815
bits = node[2][1:].split(' ')
1816
return node[0], int(bits[0]), int(bits[1])
1818
def get_method(self, version_id):
1819
"""Return compression method of specified version."""
1820
return self._get_method(self._get_node(version_id))
1822
def _get_node(self, version_id):
1824
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1826
raise RevisionNotPresent(version_id, self)
1828
def get_options(self, version_id):
1829
"""Return a list representing options.
1833
node = self._get_node(version_id)
1834
options = [self._get_method(node)]
1835
if node[2][0] == 'N':
1836
options.append('no-eol')
1839
def get_parents(self, version_id):
1840
"""Return parents of specified version ignoring ghosts."""
1841
parents = list(self.iter_parents([version_id]))
1844
raise errors.RevisionNotPresent(version_id, self)
1845
return parents[0][1]
1847
def get_parents_with_ghosts(self, version_id):
1848
"""Return parents of specified version with ghosts."""
1849
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1850
check_present=True))
1851
if not self._parents:
1853
return self._keys_to_version_ids(nodes[0][3][0])
1855
def check_versions_present(self, version_ids):
1856
"""Check that all specified versions are present."""
1857
keys = self._version_ids_to_keys(version_ids)
1858
present = self._present_keys(keys)
1859
missing = keys.difference(present)
1861
raise RevisionNotPresent(missing.pop(), self)
1863
def add_version(self, version_id, options, access_memo, parents):
1864
"""Add a version record to the index."""
1865
return self.add_versions(((version_id, options, access_memo, parents),))
1867
def add_versions(self, versions, random_id=False):
1868
"""Add multiple versions to the index.
1870
This function does not insert data into the Immutable GraphIndex
1871
backing the KnitGraphIndex, instead it prepares data for insertion by
1872
the caller and checks that it is safe to insert then calls
1873
self._add_callback with the prepared GraphIndex nodes.
1875
:param versions: a list of tuples:
1876
(version_id, options, pos, size, parents).
1877
:param random_id: If True the ids being added were randomly generated
1878
and no check for existence will be performed.
1880
if not self._add_callback:
1881
raise errors.ReadOnlyError(self)
1882
# we hope there are no repositories with inconsistent parentage
1887
for (version_id, options, access_memo, parents) in versions:
1888
index, pos, size = access_memo
1889
key = (version_id, )
1890
parents = tuple((parent, ) for parent in parents)
1891
if 'no-eol' in options:
1895
value += "%d %d" % (pos, size)
1896
if not self._deltas:
1897
if 'line-delta' in options:
1898
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1901
if 'line-delta' in options:
1902
node_refs = (parents, (parents[0],))
1904
node_refs = (parents, ())
1906
node_refs = (parents, )
1909
raise KnitCorrupt(self, "attempt to add node with parents "
1910
"in parentless index.")
1912
keys[key] = (value, node_refs)
1914
present_nodes = self._get_entries(keys)
1915
for (index, key, value, node_refs) in present_nodes:
1916
if (value, node_refs) != keys[key]:
1917
raise KnitCorrupt(self, "inconsistent details in add_versions"
1918
": %s %s" % ((value, node_refs), keys[key]))
1922
for key, (value, node_refs) in keys.iteritems():
1923
result.append((key, value, node_refs))
1925
for key, (value, node_refs) in keys.iteritems():
1926
result.append((key, value))
1927
self._add_callback(result)
1929
def _version_ids_to_keys(self, version_ids):
1930
return set((version_id, ) for version_id in version_ids)
1933
class _KnitAccess(object):
1934
"""Access to knit records in a .knit file."""
1936
def __init__(self, transport, filename, _file_mode, _dir_mode,
1937
_need_to_create, _create_parent_dir):
1938
"""Create a _KnitAccess for accessing and inserting data.
1940
:param transport: The transport the .knit is located on.
1941
:param filename: The filename of the .knit.
1943
self._transport = transport
1944
self._filename = filename
1945
self._file_mode = _file_mode
1946
self._dir_mode = _dir_mode
1947
self._need_to_create = _need_to_create
1948
self._create_parent_dir = _create_parent_dir
1950
def add_raw_records(self, sizes, raw_data):
1951
"""Add raw knit bytes to a storage area.
1953
The data is spooled to whereever the access method is storing data.
1955
:param sizes: An iterable containing the size of each raw data segment.
1956
:param raw_data: A bytestring containing the data.
1957
:return: A list of memos to retrieve the record later. Each memo is a
1958
tuple - (index, pos, length), where the index field is always None
1959
for the .knit access method.
1961
assert type(raw_data) == str, \
1962
'data must be plain bytes was %s' % type(raw_data)
1963
if not self._need_to_create:
1964
base = self._transport.append_bytes(self._filename, raw_data)
1966
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1967
create_parent_dir=self._create_parent_dir,
1968
mode=self._file_mode,
1969
dir_mode=self._dir_mode)
1970
self._need_to_create = False
1974
result.append((None, base, size))
1979
"""IFF this data access has its own storage area, initialise it.
1983
self._transport.put_bytes_non_atomic(self._filename, '',
1984
mode=self._file_mode)
1986
def open_file(self):
1987
"""IFF this data access can be represented as a single file, open it.
1989
For knits that are not mapped to a single file on disk this will
1992
:return: None or a file handle.
1995
return self._transport.get(self._filename)
2000
def get_raw_records(self, memos_for_retrieval):
2001
"""Get the raw bytes for a records.
2003
:param memos_for_retrieval: An iterable containing the (index, pos,
2004
length) memo for retrieving the bytes. The .knit method ignores
2005
the index as there is always only a single file.
2006
:return: An iterator over the bytes of the records.
2008
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
2009
for pos, data in self._transport.readv(self._filename, read_vector):
2013
class _PackAccess(object):
2014
"""Access to knit records via a collection of packs."""
2016
def __init__(self, index_to_packs, writer=None):
2017
"""Create a _PackAccess object.
2019
:param index_to_packs: A dict mapping index objects to the transport
2020
and file names for obtaining data.
2021
:param writer: A tuple (pack.ContainerWriter, write_index) which
2022
contains the pack to write, and the index that reads from it will
2026
self.container_writer = writer[0]
2027
self.write_index = writer[1]
2029
self.container_writer = None
2030
self.write_index = None
2031
self.indices = index_to_packs
2033
def add_raw_records(self, sizes, raw_data):
2034
"""Add raw knit bytes to a storage area.
2036
The data is spooled to the container writer in one bytes-record per
2039
:param sizes: An iterable containing the size of each raw data segment.
2040
:param raw_data: A bytestring containing the data.
2041
:return: A list of memos to retrieve the record later. Each memo is a
2042
tuple - (index, pos, length), where the index field is the
2043
write_index object supplied to the PackAccess object.
2045
assert type(raw_data) == str, \
2046
'data must be plain bytes was %s' % type(raw_data)
2050
p_offset, p_length = self.container_writer.add_bytes_record(
2051
raw_data[offset:offset+size], [])
2053
result.append((self.write_index, p_offset, p_length))
2057
"""Pack based knits do not get individually created."""
2059
def get_raw_records(self, memos_for_retrieval):
2060
"""Get the raw bytes for a records.
2062
:param memos_for_retrieval: An iterable containing the (index, pos,
2063
length) memo for retrieving the bytes. The Pack access method
2064
looks up the pack to use for a given record in its index_to_pack
2066
:return: An iterator over the bytes of the records.
2068
# first pass, group into same-index requests
2070
current_index = None
2071
for (index, offset, length) in memos_for_retrieval:
2072
if current_index == index:
2073
current_list.append((offset, length))
2075
if current_index is not None:
2076
request_lists.append((current_index, current_list))
2077
current_index = index
2078
current_list = [(offset, length)]
2079
# handle the last entry
2080
if current_index is not None:
2081
request_lists.append((current_index, current_list))
2082
for index, offsets in request_lists:
2083
transport, path = self.indices[index]
2084
reader = pack.make_readv_reader(transport, path, offsets)
2085
for names, read_func in reader.iter_records():
2086
yield read_func(None)
2088
def open_file(self):
2089
"""Pack based knits have no single file."""
2092
def set_writer(self, writer, index, (transport, packname)):
2093
"""Set a writer to use for adding data."""
2094
if index is not None:
2095
self.indices[index] = (transport, packname)
2096
self.container_writer = writer
2097
self.write_index = index
2100
class _StreamAccess(object):
2101
"""A Knit Access object that provides data from a datastream.
2103
It also provides a fallback to present as unannotated data, annotated data
2104
from a *backing* access object.
2106
This is triggered by a index_memo which is pointing to a different index
2107
than this was constructed with, and is used to allow extracting full
2108
unannotated texts for insertion into annotated knits.
2111
def __init__(self, reader_callable, stream_index, backing_knit,
2113
"""Create a _StreamAccess object.
2115
:param reader_callable: The reader_callable from the datastream.
2116
This is called to buffer all the data immediately, for
2118
:param stream_index: The index the data stream this provides access to
2119
which will be present in native index_memo's.
2120
:param backing_knit: The knit object that will provide access to
2121
annotated texts which are not available in the stream, so as to
2122
create unannotated texts.
2123
:param orig_factory: The original content factory used to generate the
2124
stream. This is used for checking whether the thunk code for
2125
supporting _copy_texts will generate the correct form of data.
2127
self.data = reader_callable(None)
2128
self.stream_index = stream_index
2129
self.backing_knit = backing_knit
2130
self.orig_factory = orig_factory
2132
def get_raw_records(self, memos_for_retrieval):
2133
"""Get the raw bytes for a records.
2135
:param memos_for_retrieval: An iterable containing the (thunk_flag,
2136
index, start, end) memo for retrieving the bytes.
2137
:return: An iterator over the bytes of the records.
2139
# use a generator for memory friendliness
2140
for thunk_flag, version_id, start, end in memos_for_retrieval:
2141
if version_id is self.stream_index:
2142
yield self.data[start:end]
2144
# we have been asked to thunk. This thunking only occurs when
2145
# we are obtaining plain texts from an annotated backing knit
2146
# so that _copy_texts will work.
2147
# We could improve performance here by scanning for where we need
2148
# to do this and using get_line_list, then interleaving the output
2149
# as desired. However, for now, this is sufficient.
2150
if self.orig_factory.__class__ != KnitPlainFactory:
2151
raise errors.KnitCorrupt(
2152
self, 'Bad thunk request %r' % version_id)
2153
lines = self.backing_knit.get_lines(version_id)
2154
line_bytes = ''.join(lines)
2155
digest = sha_string(line_bytes)
2157
if lines[-1][-1] != '\n':
2158
lines[-1] = lines[-1] + '\n'
2160
orig_options = list(self.backing_knit._index.get_options(version_id))
2161
if 'fulltext' not in orig_options:
2162
if 'line-delta' not in orig_options:
2163
raise errors.KnitCorrupt(self,
2164
'Unknown compression method %r' % orig_options)
2165
orig_options.remove('line-delta')
2166
orig_options.append('fulltext')
2167
# We want plain data, because we expect to thunk only to allow text
2169
size, bytes = self.backing_knit._data._record_to_data(version_id,
2170
digest, lines, line_bytes)
2174
class _StreamIndex(object):
2175
"""A Knit Index object that uses the data map from a datastream."""
2177
def __init__(self, data_list):
2178
"""Create a _StreamIndex object.
2180
:param data_list: The data_list from the datastream.
2182
self.data_list = data_list
2183
self._by_version = {}
2185
for key, options, length, parents in data_list:
2186
self._by_version[key] = options, (pos, pos + length), parents
2189
def get_ancestry(self, versions, topo_sorted):
2190
"""Get an ancestry list for versions."""
2192
# Not needed for basic joins
2193
raise NotImplementedError(self.get_ancestry)
2194
# get a graph of all the mentioned versions:
2195
# Little ugly - basically copied from KnitIndex, but don't want to
2196
# accidentally incorporate too much of that index's code.
2198
pending = set(versions)
2199
cache = self._by_version
2201
version = pending.pop()
2204
parents = [p for p in cache[version][2] if p in cache]
2206
raise RevisionNotPresent(version, self)
2207
# if not completed and not a ghost
2208
pending.update([p for p in parents if p not in ancestry])
2209
ancestry.add(version)
2210
return list(ancestry)
2212
def get_build_details(self, version_ids):
2213
"""Get the method, index_memo and compression parent for version_ids.
2215
:param version_ids: An iterable of version_ids.
2216
:return: A dict of version_id:(method, index_memo, compression_parent).
2219
for version_id in version_ids:
2220
method = self.get_method(version_id)
2221
if method == 'fulltext':
2222
compression_parent = None
2224
compression_parent = self.get_parents_with_ghosts(version_id)[0]
2225
index_memo = self.get_position(version_id)
2226
result[version_id] = (method, index_memo, compression_parent)
2229
def get_method(self, version_id):
2230
"""Return compression method of specified version."""
2232
options = self._by_version[version_id][0]
2234
# Strictly speaking this should check in the backing knit, but
2235
# until we have a test to discriminate, this will do.
2237
if 'fulltext' in options:
2239
elif 'line-delta' in options:
2242
raise errors.KnitIndexUnknownMethod(self, options)
2244
def get_options(self, version_id):
2245
"""Return a list representing options.
2249
return self._by_version[version_id][0]
2251
def get_parents_with_ghosts(self, version_id):
2252
"""Return parents of specified version with ghosts."""
2253
return self._by_version[version_id][2]
2255
def get_position(self, version_id):
2256
"""Return details needed to access the version.
2258
_StreamAccess has the data as a big array, so we return slice
2259
coordinates into that (as index_memo's are opaque outside the
2260
index and matching access class).
2262
:return: a tuple (thunk_flag, index, start, end). If thunk_flag is
2263
False, index will be self, otherwise it will be a version id.
2266
start, end = self._by_version[version_id][1]
2267
return False, self, start, end
2269
# Signal to the access object to handle this from the backing knit.
2270
return (True, version_id, None, None)
2272
def get_versions(self):
2273
"""Get all the versions in the stream."""
2274
return self._by_version.keys()
2276
def iter_parents(self, version_ids):
2277
"""Iterate through the parents for many version ids.
2279
:param version_ids: An iterable yielding version_ids.
2280
:return: An iterator that yields (version_id, parents). Requested
2281
version_ids not present in the versioned file are simply skipped.
2282
The order is undefined, allowing for different optimisations in
2283
the underlying implementation.
2286
for version in version_ids:
2288
result.append((version, self._by_version[version][2]))
2294
class _KnitData(object):
2295
"""Manage extraction of data from a KnitAccess, caching and decompressing.
2297
The KnitData class provides the logic for parsing and using knit records,
2298
making use of an access method for the low level read and write operations.
2301
def __init__(self, access):
2302
"""Create a KnitData object.
2304
:param access: The access method to use. Access methods such as
2305
_KnitAccess manage the insertion of raw records and the subsequent
2306
retrieval of the same.
2308
self._access = access
2309
self._checked = False
2310
# TODO: jam 20060713 conceptually, this could spill to disk
2311
# if the cached size gets larger than a certain amount
2312
# but it complicates the model a bit, so for now just use
2313
# a simple dictionary
2315
self._do_cache = False
2317
def enable_cache(self):
2318
"""Enable caching of reads."""
2319
self._do_cache = True
2321
def clear_cache(self):
2322
"""Clear the record cache."""
2323
self._do_cache = False
2326
def _open_file(self):
2327
return self._access.open_file()
2329
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2330
"""Convert version_id, digest, lines into a raw data block.
2332
:param dense_lines: The bytes of lines but in a denser form. For
2333
instance, if lines is a list of 1000 bytestrings each ending in \n,
2334
dense_lines may be a list with one line in it, containing all the
2335
1000's lines and their \n's. Using dense_lines if it is already
2336
known is a win because the string join to create bytes in this
2337
function spends less time resizing the final string.
2338
:return: (len, a StringIO instance with the raw data ready to read.)
2340
# Note: using a string copy here increases memory pressure with e.g.
2341
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2342
# when doing the initial commit of a mozilla tree. RBC 20070921
2343
bytes = ''.join(chain(
2344
["version %s %d %s\n" % (version_id,
2347
dense_lines or lines,
2348
["end %s\n" % version_id]))
2349
assert bytes.__class__ == str
2350
compressed_bytes = bytes_to_gzip(bytes)
2351
return len(compressed_bytes), compressed_bytes
2353
def add_raw_records(self, sizes, raw_data):
2354
"""Append a prepared record to the data file.
2356
:param sizes: An iterable containing the size of each raw data segment.
2357
:param raw_data: A bytestring containing the data.
2358
:return: a list of index data for the way the data was stored.
2359
See the access method add_raw_records documentation for more
2362
return self._access.add_raw_records(sizes, raw_data)
2364
def _parse_record_header(self, version_id, raw_data):
2365
"""Parse a record header for consistency.
2367
:return: the header and the decompressor stream.
2368
as (stream, header_record)
2370
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2372
rec = self._check_header(version_id, df.readline())
2373
except Exception, e:
2374
raise KnitCorrupt(self._access,
2375
"While reading {%s} got %s(%s)"
2376
% (version_id, e.__class__.__name__, str(e)))
2379
def _check_header(self, version_id, line):
2382
raise KnitCorrupt(self._access,
2383
'unexpected number of elements in record header')
2384
if rec[1] != version_id:
2385
raise KnitCorrupt(self._access,
2386
'unexpected version, wanted %r, got %r'
2387
% (version_id, rec[1]))
2390
def _parse_record(self, version_id, data):
2392
# 4168 calls in 2880 217 internal
2393
# 4168 calls to _parse_record_header in 2121
2394
# 4168 calls to readlines in 330
2395
df = GzipFile(mode='rb', fileobj=StringIO(data))
2398
record_contents = df.readlines()
2399
except Exception, e:
2400
raise KnitCorrupt(self._access,
2401
"While reading {%s} got %s(%s)"
2402
% (version_id, e.__class__.__name__, str(e)))
2403
header = record_contents.pop(0)
2404
rec = self._check_header(version_id, header)
2406
last_line = record_contents.pop()
2407
if len(record_contents) != int(rec[2]):
2408
raise KnitCorrupt(self._access,
2409
'incorrect number of lines %s != %s'
2411
% (len(record_contents), int(rec[2]),
2413
if last_line != 'end %s\n' % rec[1]:
2414
raise KnitCorrupt(self._access,
2415
'unexpected version end line %r, wanted %r'
2416
% (last_line, version_id))
2418
return record_contents, rec[3]
2420
def read_records_iter_raw(self, records):
2421
"""Read text records from data file and yield raw data.
2423
This unpacks enough of the text record to validate the id is
2424
as expected but thats all.
2426
# setup an iterator of the external records:
2427
# uses readv so nice and fast we hope.
2429
# grab the disk data needed.
2431
# Don't check _cache if it is empty
2432
needed_offsets = [index_memo for version_id, index_memo
2434
if version_id not in self._cache]
2436
needed_offsets = [index_memo for version_id, index_memo
2439
raw_records = self._access.get_raw_records(needed_offsets)
2441
for version_id, index_memo in records:
2442
if version_id in self._cache:
2443
# This data has already been validated
2444
data = self._cache[version_id]
2446
data = raw_records.next()
2448
self._cache[version_id] = data
2450
# validate the header
2451
df, rec = self._parse_record_header(version_id, data)
2453
yield version_id, data
2455
def read_records_iter(self, records):
2456
"""Read text records from data file and yield result.
2458
The result will be returned in whatever is the fastest to read.
2459
Not by the order requested. Also, multiple requests for the same
2460
record will only yield 1 response.
2461
:param records: A list of (version_id, pos, len) entries
2462
:return: Yields (version_id, contents, digest) in the order
2463
read, not the order requested
2469
# Skip records we have alread seen
2470
yielded_records = set()
2471
needed_records = set()
2472
for record in records:
2473
if record[0] in self._cache:
2474
if record[0] in yielded_records:
2476
yielded_records.add(record[0])
2477
data = self._cache[record[0]]
2478
content, digest = self._parse_record(record[0], data)
2479
yield (record[0], content, digest)
2481
needed_records.add(record)
2482
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2484
needed_records = sorted(set(records), key=operator.itemgetter(1))
2486
if not needed_records:
2489
# The transport optimizes the fetching as well
2490
# (ie, reads continuous ranges.)
2491
raw_data = self._access.get_raw_records(
2492
[index_memo for version_id, index_memo in needed_records])
2494
for (version_id, index_memo), data in \
2495
izip(iter(needed_records), raw_data):
2496
content, digest = self._parse_record(version_id, data)
2498
self._cache[version_id] = data
2499
yield version_id, content, digest
2501
def read_records(self, records):
2502
"""Read records into a dictionary."""
2504
for record_id, content, digest in \
2505
self.read_records_iter(records):
2506
components[record_id] = (content, digest)
2510
class InterKnit(InterVersionedFile):
2511
"""Optimised code paths for knit to knit operations."""
2513
_matching_file_from_factory = KnitVersionedFile
2514
_matching_file_to_factory = KnitVersionedFile
2517
def is_compatible(source, target):
2518
"""Be compatible with knits. """
2520
return (isinstance(source, KnitVersionedFile) and
2521
isinstance(target, KnitVersionedFile))
2522
except AttributeError:
2525
def _copy_texts(self, pb, msg, version_ids, ignore_missing=False):
2526
"""Copy texts to the target by extracting and adding them one by one.
2528
see join() for the parameter definitions.
2530
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2531
graph = self.source.get_graph(version_ids)
2532
order = topo_sort(graph.items())
2534
def size_of_content(content):
2535
return sum(len(line) for line in content.text())
2536
# Cache at most 10MB of parent texts
2537
parent_cache = lru_cache.LRUSizeCache(max_size=10*1024*1024,
2538
compute_size=size_of_content)
2539
# TODO: jam 20071116 It would be nice to have a streaming interface to
2540
# get multiple texts from a source. The source could be smarter
2541
# about how it handled intermediate stages.
2542
# get_line_list() or make_mpdiffs() seem like a possibility, but
2543
# at the moment they extract all full texts into memory, which
2544
# causes us to store more than our 3x fulltext goal.
2545
# Repository.iter_files_bytes() may be another possibility
2546
to_process = [version for version in order
2547
if version not in self.target]
2548
total = len(to_process)
2549
pb = ui.ui_factory.nested_progress_bar()
2551
for index, version in enumerate(to_process):
2552
pb.update('Converting versioned data', index, total)
2553
sha1, num_bytes, parent_text = self.target.add_lines(version,
2554
self.source.get_parents_with_ghosts(version),
2555
self.source.get_lines(version),
2556
parent_texts=parent_cache)
2557
parent_cache[version] = parent_text
2562
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2563
"""See InterVersionedFile.join."""
2564
assert isinstance(self.source, KnitVersionedFile)
2565
assert isinstance(self.target, KnitVersionedFile)
2567
# If the source and target are mismatched w.r.t. annotations vs
2568
# plain, the data needs to be converted accordingly
2569
if self.source.factory.annotated == self.target.factory.annotated:
2571
elif self.source.factory.annotated:
2572
converter = self._anno_to_plain_converter
2574
# We're converting from a plain to an annotated knit. Copy them
2575
# across by full texts.
2576
return self._copy_texts(pb, msg, version_ids, ignore_missing)
2578
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2582
pb = ui.ui_factory.nested_progress_bar()
2584
version_ids = list(version_ids)
2585
if None in version_ids:
2586
version_ids.remove(None)
2588
self.source_ancestry = set(self.source.get_ancestry(version_ids,
2590
this_versions = set(self.target._index.get_versions())
2591
# XXX: For efficiency we should not look at the whole index,
2592
# we only need to consider the referenced revisions - they
2593
# must all be present, or the method must be full-text.
2594
# TODO, RBC 20070919
2595
needed_versions = self.source_ancestry - this_versions
2597
if not needed_versions:
2599
full_list = topo_sort(self.source.get_graph())
2601
version_list = [i for i in full_list if (not self.target.has_version(i)
2602
and i in needed_versions)]
2606
copy_queue_records = []
2608
for version_id in version_list:
2609
options = self.source._index.get_options(version_id)
2610
parents = self.source._index.get_parents_with_ghosts(version_id)
2611
# check that its will be a consistent copy:
2612
for parent in parents:
2613
# if source has the parent, we must :
2614
# * already have it or
2615
# * have it scheduled already
2616
# otherwise we don't care
2617
assert (self.target.has_version(parent) or
2618
parent in copy_set or
2619
not self.source.has_version(parent))
2620
index_memo = self.source._index.get_position(version_id)
2621
copy_queue_records.append((version_id, index_memo))
2622
copy_queue.append((version_id, options, parents))
2623
copy_set.add(version_id)
2625
# data suck the join:
2627
total = len(version_list)
2630
for (version_id, raw_data), \
2631
(version_id2, options, parents) in \
2632
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2634
assert version_id == version_id2, 'logic error, inconsistent results'
2636
pb.update("Joining knit", count, total)
2638
size, raw_data = converter(raw_data, version_id, options,
2641
size = len(raw_data)
2642
raw_records.append((version_id, options, parents, size))
2643
raw_datum.append(raw_data)
2644
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2649
def _anno_to_plain_converter(self, raw_data, version_id, options,
2651
"""Convert annotated content to plain content."""
2652
data, digest = self.source._data._parse_record(version_id, raw_data)
2653
if 'fulltext' in options:
2654
content = self.source.factory.parse_fulltext(data, version_id)
2655
lines = self.target.factory.lower_fulltext(content)
2657
delta = self.source.factory.parse_line_delta(data, version_id,
2659
lines = self.target.factory.lower_line_delta(delta)
2660
return self.target._data._record_to_data(version_id, digest, lines)
2663
InterVersionedFile.register_optimiser(InterKnit)
2666
class WeaveToKnit(InterVersionedFile):
2667
"""Optimised code paths for weave to knit operations."""
2669
_matching_file_from_factory = bzrlib.weave.WeaveFile
2670
_matching_file_to_factory = KnitVersionedFile
2673
def is_compatible(source, target):
2674
"""Be compatible with weaves to knits."""
2676
return (isinstance(source, bzrlib.weave.Weave) and
2677
isinstance(target, KnitVersionedFile))
2678
except AttributeError:
2681
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2682
"""See InterVersionedFile.join."""
2683
assert isinstance(self.source, bzrlib.weave.Weave)
2684
assert isinstance(self.target, KnitVersionedFile)
2686
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2691
pb = ui.ui_factory.nested_progress_bar()
2693
version_ids = list(version_ids)
2695
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2696
this_versions = set(self.target._index.get_versions())
2697
needed_versions = self.source_ancestry - this_versions
2699
if not needed_versions:
2701
full_list = topo_sort(self.source.get_graph())
2703
version_list = [i for i in full_list if (not self.target.has_version(i)
2704
and i in needed_versions)]
2708
total = len(version_list)
2709
for version_id in version_list:
2710
pb.update("Converting to knit", count, total)
2711
parents = self.source.get_parents(version_id)
2712
# check that its will be a consistent copy:
2713
for parent in parents:
2714
# if source has the parent, we must already have it
2715
assert (self.target.has_version(parent))
2716
self.target.add_lines(
2717
version_id, parents, self.source.get_lines(version_id))
2724
InterVersionedFile.register_optimiser(WeaveToKnit)
2727
# Deprecated, use PatienceSequenceMatcher instead
2728
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2731
def annotate_knit(knit, revision_id):
2732
"""Annotate a knit with no cached annotations.
2734
This implementation is for knits with no cached annotations.
2735
It will work for knits with cached annotations, but this is not
2738
ancestry = knit.get_ancestry(revision_id)
2739
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2741
for candidate in ancestry:
2742
if candidate in annotations:
2744
parents = knit.get_parents(candidate)
2745
if len(parents) == 0:
2747
elif knit._index.get_method(candidate) != 'line-delta':
2750
parent, sha1, noeol, delta = knit.get_delta(candidate)
2751
blocks = KnitContent.get_line_delta_blocks(delta,
2752
fulltext[parents[0]], fulltext[candidate])
2753
annotations[candidate] = list(annotate.reannotate([annotations[p]
2754
for p in parents], fulltext[candidate], candidate, blocks))
2755
return iter(annotations[revision_id])
2759
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2761
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
