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(), """
92
from bzrlib.errors import (
98
KnitDataStreamIncompatible,
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 line_delta_iter(self, new_lines):
142
"""Generate line-based delta from this content to new_lines."""
143
new_texts = new_lines.text()
144
old_texts = self.text()
145
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
146
for tag, i1, i2, j1, j2 in s.get_opcodes():
149
# ofrom, oto, length, data
150
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
152
def line_delta(self, new_lines):
153
return list(self.line_delta_iter(new_lines))
156
def get_line_delta_blocks(knit_delta, source, target):
157
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
158
target_len = len(target)
161
for s_begin, s_end, t_len, new_text in knit_delta:
162
true_n = s_begin - s_pos
165
# knit deltas do not provide reliable info about whether the
166
# last line of a file matches, due to eol handling.
167
if source[s_pos + n -1] != target[t_pos + n -1]:
170
yield s_pos, t_pos, n
171
t_pos += t_len + true_n
173
n = target_len - t_pos
175
if source[s_pos + n -1] != target[t_pos + n -1]:
178
yield s_pos, t_pos, n
179
yield s_pos + (target_len - t_pos), target_len, 0
182
class AnnotatedKnitContent(KnitContent):
183
"""Annotated content."""
185
def __init__(self, lines):
188
def annotate_iter(self):
189
"""Yield tuples of (origin, text) for each content line."""
190
return iter(self._lines)
192
def strip_last_line_newline(self):
193
line = self._lines[-1][1].rstrip('\n')
194
self._lines[-1] = (self._lines[-1][0], line)
197
return [text for origin, text in self._lines]
200
return AnnotatedKnitContent(self._lines[:])
203
class PlainKnitContent(KnitContent):
204
"""Unannotated content.
206
When annotate[_iter] is called on this content, the same version is reported
207
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
211
def __init__(self, lines, version_id):
213
self._version_id = version_id
215
def annotate_iter(self):
216
"""Yield tuples of (origin, text) for each content line."""
217
for line in self._lines:
218
yield self._version_id, line
221
return PlainKnitContent(self._lines[:], self._version_id)
223
def strip_last_line_newline(self):
224
self._lines[-1] = self._lines[-1].rstrip('\n')
230
class KnitAnnotateFactory(object):
231
"""Factory for creating annotated Content objects."""
235
def make(self, lines, version_id):
236
num_lines = len(lines)
237
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
239
def parse_fulltext(self, content, version_id):
240
"""Convert fulltext to internal representation
242
fulltext content is of the format
243
revid(utf8) plaintext\n
244
internal representation is of the format:
247
# TODO: jam 20070209 The tests expect this to be returned as tuples,
248
# but the code itself doesn't really depend on that.
249
# Figure out a way to not require the overhead of turning the
250
# list back into tuples.
251
lines = [tuple(line.split(' ', 1)) for line in content]
252
return AnnotatedKnitContent(lines)
254
def parse_line_delta_iter(self, lines):
255
return iter(self.parse_line_delta(lines))
257
def parse_line_delta(self, lines, version_id):
258
"""Convert a line based delta into internal representation.
260
line delta is in the form of:
261
intstart intend intcount
263
revid(utf8) newline\n
264
internal representation is
265
(start, end, count, [1..count tuples (revid, newline)])
272
def cache_and_return(line):
273
origin, text = line.split(' ', 1)
274
return cache.setdefault(origin, origin), text
276
# walk through the lines parsing.
278
start, end, count = [int(n) for n in header.split(',')]
279
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
280
result.append((start, end, count, contents))
283
def get_fulltext_content(self, lines):
284
"""Extract just the content lines from a fulltext."""
285
return (line.split(' ', 1)[1] for line in lines)
287
def get_linedelta_content(self, lines):
288
"""Extract just the content from a line delta.
290
This doesn't return all of the extra information stored in a delta.
291
Only the actual content lines.
296
header = header.split(',')
297
count = int(header[2])
298
for i in xrange(count):
299
origin, text = next().split(' ', 1)
302
def lower_fulltext(self, content):
303
"""convert a fulltext content record into a serializable form.
305
see parse_fulltext which this inverts.
307
# TODO: jam 20070209 We only do the caching thing to make sure that
308
# the origin is a valid utf-8 line, eventually we could remove it
309
return ['%s %s' % (o, t) for o, t in content._lines]
311
def lower_line_delta(self, delta):
312
"""convert a delta into a serializable form.
314
See parse_line_delta which this inverts.
316
# TODO: jam 20070209 We only do the caching thing to make sure that
317
# the origin is a valid utf-8 line, eventually we could remove it
319
for start, end, c, lines in delta:
320
out.append('%d,%d,%d\n' % (start, end, c))
321
out.extend(origin + ' ' + text
322
for origin, text in lines)
325
def annotate_iter(self, knit, version_id):
326
content = knit._get_content(version_id)
327
return content.annotate_iter()
330
class KnitPlainFactory(object):
331
"""Factory for creating plain Content objects."""
335
def make(self, lines, version_id):
336
return PlainKnitContent(lines, version_id)
338
def parse_fulltext(self, content, version_id):
339
"""This parses an unannotated fulltext.
341
Note that this is not a noop - the internal representation
342
has (versionid, line) - its just a constant versionid.
344
return self.make(content, version_id)
346
def parse_line_delta_iter(self, lines, version_id):
348
num_lines = len(lines)
349
while cur < num_lines:
352
start, end, c = [int(n) for n in header.split(',')]
353
yield start, end, c, lines[cur:cur+c]
356
def parse_line_delta(self, lines, version_id):
357
return list(self.parse_line_delta_iter(lines, version_id))
359
def get_fulltext_content(self, lines):
360
"""Extract just the content lines from a fulltext."""
363
def get_linedelta_content(self, lines):
364
"""Extract just the content from a line delta.
366
This doesn't return all of the extra information stored in a delta.
367
Only the actual content lines.
372
header = header.split(',')
373
count = int(header[2])
374
for i in xrange(count):
377
def lower_fulltext(self, content):
378
return content.text()
380
def lower_line_delta(self, delta):
382
for start, end, c, lines in delta:
383
out.append('%d,%d,%d\n' % (start, end, c))
387
def annotate_iter(self, knit, version_id):
388
return annotate_knit(knit, version_id)
391
def make_empty_knit(transport, relpath):
392
"""Construct a empty knit at the specified location."""
393
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
396
class KnitVersionedFile(VersionedFile):
397
"""Weave-like structure with faster random access.
399
A knit stores a number of texts and a summary of the relationships
400
between them. Texts are identified by a string version-id. Texts
401
are normally stored and retrieved as a series of lines, but can
402
also be passed as single strings.
404
Lines are stored with the trailing newline (if any) included, to
405
avoid special cases for files with no final newline. Lines are
406
composed of 8-bit characters, not unicode. The combination of
407
these approaches should mean any 'binary' file can be safely
408
stored and retrieved.
411
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
412
factory=None, basis_knit=DEPRECATED_PARAMETER, delta=True,
413
create=False, create_parent_dir=False, delay_create=False,
414
dir_mode=None, index=None, access_method=None):
415
"""Construct a knit at location specified by relpath.
417
:param create: If not True, only open an existing knit.
418
:param create_parent_dir: If True, create the parent directory if
419
creating the file fails. (This is used for stores with
420
hash-prefixes that may not exist yet)
421
:param delay_create: The calling code is aware that the knit won't
422
actually be created until the first data is stored.
423
:param index: An index to use for the knit.
425
if deprecated_passed(basis_knit):
426
warnings.warn("KnitVersionedFile.__(): The basis_knit parameter is"
427
" deprecated as of bzr 0.9.",
428
DeprecationWarning, stacklevel=2)
429
if access_mode is None:
431
super(KnitVersionedFile, self).__init__(access_mode)
432
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
433
self.transport = transport
434
self.filename = relpath
435
self.factory = factory or KnitAnnotateFactory()
436
self.writable = (access_mode == 'w')
439
self._max_delta_chain = 200
442
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
443
access_mode, create=create, file_mode=file_mode,
444
create_parent_dir=create_parent_dir, delay_create=delay_create,
448
if access_method is None:
449
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
450
((create and not len(self)) and delay_create), create_parent_dir)
452
_access = access_method
453
if create and not len(self) and not delay_create:
455
self._data = _KnitData(_access)
458
return '%s(%s)' % (self.__class__.__name__,
459
self.transport.abspath(self.filename))
461
def _check_should_delta(self, first_parents):
462
"""Iterate back through the parent listing, looking for a fulltext.
464
This is used when we want to decide whether to add a delta or a new
465
fulltext. It searches for _max_delta_chain parents. When it finds a
466
fulltext parent, it sees if the total size of the deltas leading up to
467
it is large enough to indicate that we want a new full text anyway.
469
Return True if we should create a new delta, False if we should use a
474
delta_parents = first_parents
475
for count in xrange(self._max_delta_chain):
476
parent = delta_parents[0]
477
method = self._index.get_method(parent)
478
index, pos, size = self._index.get_position(parent)
479
if method == 'fulltext':
483
delta_parents = self._index.get_parents(parent)
485
# We couldn't find a fulltext, so we must create a new one
488
return fulltext_size > delta_size
490
def _add_raw_records(self, records, data):
491
"""Add all the records 'records' with data pre-joined in 'data'.
493
:param records: A list of tuples(version_id, options, parents, size).
494
:param data: The data for the records. When it is written, the records
495
are adjusted to have pos pointing into data by the sum of
496
the preceding records sizes.
499
raw_record_sizes = [record[3] for record in records]
500
positions = self._data.add_raw_records(raw_record_sizes, data)
503
for (version_id, options, parents, size), access_memo in zip(
505
index_entries.append((version_id, options, access_memo, parents))
506
if self._data._do_cache:
507
self._data._cache[version_id] = data[offset:offset+size]
509
self._index.add_versions(index_entries)
511
def enable_cache(self):
512
"""Start caching data for this knit"""
513
self._data.enable_cache()
515
def clear_cache(self):
516
"""Clear the data cache only."""
517
self._data.clear_cache()
519
def copy_to(self, name, transport):
520
"""See VersionedFile.copy_to()."""
521
# copy the current index to a temp index to avoid racing with local
523
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
524
self.transport.get(self._index._filename))
526
f = self._data._open_file()
528
transport.put_file(name + DATA_SUFFIX, f)
531
# move the copied index into place
532
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
534
def create_empty(self, name, transport, mode=None):
535
return KnitVersionedFile(name, transport, factory=self.factory,
536
delta=self.delta, create=True)
538
def get_data_stream(self, required_versions):
539
"""Get a data stream for the specified versions.
541
Versions may be returned in any order, not necessarily the order
544
:param required_versions: The exact set of versions to be extracted.
545
Unlike some other knit methods, this is not used to generate a
546
transitive closure, rather it is used precisely as given.
548
:returns: format_signature, list of (version, options, length, parents),
551
if not isinstance(required_versions, set):
552
required_versions = set(required_versions)
553
# we don't care about inclusions, the caller cares.
554
# but we need to setup a list of records to visit.
555
for version_id in required_versions:
556
if not self.has_version(version_id):
557
raise RevisionNotPresent(version_id, self.filename)
558
# Pick the desired versions out of the index in oldest-to-newest order
560
for version_id in self.versions():
561
if version_id in required_versions:
562
version_list.append(version_id)
564
# create the list of version information for the result
565
copy_queue_records = []
567
result_version_list = []
568
for version_id in version_list:
569
options = self._index.get_options(version_id)
570
parents = self._index.get_parents_with_ghosts(version_id)
571
index_memo = self._index.get_position(version_id)
572
copy_queue_records.append((version_id, index_memo))
573
none, data_pos, data_size = index_memo
574
copy_set.add(version_id)
575
# version, options, length, parents
576
result_version_list.append((version_id, options, data_size,
579
# Read the compressed record data.
581
# From here down to the return should really be logic in the returned
582
# callable -- in a class that adapts read_records_iter_raw to read
585
for (version_id, raw_data), \
586
(version_id2, options, _, parents) in \
587
izip(self._data.read_records_iter_raw(copy_queue_records),
588
result_version_list):
589
assert version_id == version_id2, 'logic error, inconsistent results'
590
raw_datum.append(raw_data)
591
pseudo_file = StringIO(''.join(raw_datum))
594
return pseudo_file.read()
596
return pseudo_file.read(length)
597
return (self.get_format_signature(), result_version_list, read)
599
def _extract_blocks(self, version_id, source, target):
600
if self._index.get_method(version_id) != 'line-delta':
602
parent, sha1, noeol, delta = self.get_delta(version_id)
603
return KnitContent.get_line_delta_blocks(delta, source, target)
605
def get_delta(self, version_id):
606
"""Get a delta for constructing version from some other version."""
607
self.check_not_reserved_id(version_id)
608
parents = self.get_parents(version_id)
613
index_memo = self._index.get_position(version_id)
614
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
615
noeol = 'no-eol' in self._index.get_options(version_id)
616
if 'fulltext' == self._index.get_method(version_id):
617
new_content = self.factory.parse_fulltext(data, version_id)
618
if parent is not None:
619
reference_content = self._get_content(parent)
620
old_texts = reference_content.text()
623
new_texts = new_content.text()
624
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
626
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
628
delta = self.factory.parse_line_delta(data, version_id)
629
return parent, sha1, noeol, delta
631
def get_format_signature(self):
632
"""See VersionedFile.get_format_signature()."""
633
if self.factory.annotated:
634
annotated_part = "annotated"
636
annotated_part = "plain"
637
return "knit-%s" % (annotated_part,)
639
def get_graph_with_ghosts(self):
640
"""See VersionedFile.get_graph_with_ghosts()."""
641
graph_items = self._index.get_graph()
642
return dict(graph_items)
644
def get_sha1(self, version_id):
645
return self.get_sha1s([version_id])[0]
647
def get_sha1s(self, version_ids):
648
"""See VersionedFile.get_sha1()."""
649
record_map = self._get_record_map(version_ids)
650
# record entry 2 is the 'digest'.
651
return [record_map[v][2] for v in version_ids]
655
"""See VersionedFile.get_suffixes()."""
656
return [DATA_SUFFIX, INDEX_SUFFIX]
658
def has_ghost(self, version_id):
659
"""True if there is a ghost reference in the file to version_id."""
661
if self.has_version(version_id):
663
# optimisable if needed by memoising the _ghosts set.
664
items = self._index.get_graph()
665
for node, parents in items:
666
for parent in parents:
667
if parent not in self._index._cache:
668
if parent == version_id:
672
def insert_data_stream(self, (format, data_list, reader_callable)):
673
"""Insert knit records from a data stream into this knit.
675
If a version in the stream is already present in this knit, it will not
676
be inserted a second time. It will be checked for consistency with the
677
stored version however, and may cause a KnitCorrupt error to be raised
678
if the data in the stream disagrees with the already stored data.
680
:seealso: get_data_stream
682
if format != self.get_format_signature():
683
trace.mutter('incompatible format signature inserting to %r', self)
684
raise KnitDataStreamIncompatible(
685
format, self.get_format_signature())
687
for version_id, options, length, parents in data_list:
688
if self.has_version(version_id):
689
# First check: the list of parents.
690
my_parents = self.get_parents_with_ghosts(version_id)
691
if my_parents != parents:
692
# XXX: KnitCorrupt is not quite the right exception here.
695
'parents list %r from data stream does not match '
696
'already recorded parents %r for %s'
697
% (parents, my_parents, version_id))
699
# Also check the SHA-1 of the fulltext this content will
701
raw_data = reader_callable(length)
702
my_fulltext_sha1 = self.get_sha1(version_id)
703
df, rec = self._data._parse_record_header(version_id, raw_data)
704
stream_fulltext_sha1 = rec[3]
705
if my_fulltext_sha1 != stream_fulltext_sha1:
706
# Actually, we don't know if it's this knit that's corrupt,
707
# or the data stream we're trying to insert.
709
self.filename, 'sha-1 does not match %s' % version_id)
711
self._add_raw_records(
712
[(version_id, options, parents, length)],
713
reader_callable(length))
716
"""See VersionedFile.versions."""
717
if 'evil' in debug.debug_flags:
718
trace.mutter_callsite(2, "versions scales with size of history")
719
return self._index.get_versions()
721
def has_version(self, version_id):
722
"""See VersionedFile.has_version."""
723
if 'evil' in debug.debug_flags:
724
trace.mutter_callsite(2, "has_version is a LBYL scenario")
725
return self._index.has_version(version_id)
727
__contains__ = has_version
729
def _merge_annotations(self, content, parents, parent_texts={},
730
delta=None, annotated=None,
731
left_matching_blocks=None):
732
"""Merge annotations for content. This is done by comparing
733
the annotations based on changed to the text.
735
if left_matching_blocks is not None:
736
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
740
for parent_id in parents:
741
merge_content = self._get_content(parent_id, parent_texts)
742
if (parent_id == parents[0] and delta_seq is not None):
745
seq = patiencediff.PatienceSequenceMatcher(
746
None, merge_content.text(), content.text())
747
for i, j, n in seq.get_matching_blocks():
750
# this appears to copy (origin, text) pairs across to the
751
# new content for any line that matches the last-checked
753
content._lines[j:j+n] = merge_content._lines[i:i+n]
755
if delta_seq is None:
756
reference_content = self._get_content(parents[0], parent_texts)
757
new_texts = content.text()
758
old_texts = reference_content.text()
759
delta_seq = patiencediff.PatienceSequenceMatcher(
760
None, old_texts, new_texts)
761
return self._make_line_delta(delta_seq, content)
763
def _make_line_delta(self, delta_seq, new_content):
764
"""Generate a line delta from delta_seq and new_content."""
766
for op in delta_seq.get_opcodes():
769
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
772
def _get_components_positions(self, version_ids):
773
"""Produce a map of position data for the components of versions.
775
This data is intended to be used for retrieving the knit records.
777
A dict of version_id to (method, data_pos, data_size, next) is
779
method is the way referenced data should be applied.
780
data_pos is the position of the data in the knit.
781
data_size is the size of the data in the knit.
782
next is the build-parent of the version, or None for fulltexts.
785
for version_id in version_ids:
788
while cursor is not None and cursor not in component_data:
789
method = self._index.get_method(cursor)
790
if method == 'fulltext':
793
next = self.get_parents(cursor)[0]
794
index_memo = self._index.get_position(cursor)
795
component_data[cursor] = (method, index_memo, next)
797
return component_data
799
def _get_content(self, version_id, parent_texts={}):
800
"""Returns a content object that makes up the specified
802
cached_version = parent_texts.get(version_id, None)
803
if cached_version is not None:
804
if not self.has_version(version_id):
805
raise RevisionNotPresent(version_id, self.filename)
806
return cached_version
808
text_map, contents_map = self._get_content_maps([version_id])
809
return contents_map[version_id]
811
def _check_versions_present(self, version_ids):
812
"""Check that all specified versions are present."""
813
self._index.check_versions_present(version_ids)
815
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
816
nostore_sha, random_id, check_content):
817
"""See VersionedFile.add_lines_with_ghosts()."""
818
self._check_add(version_id, lines, random_id, check_content)
819
return self._add(version_id, lines, parents, self.delta,
820
parent_texts, None, nostore_sha, random_id)
822
def _add_lines(self, version_id, parents, lines, parent_texts,
823
left_matching_blocks, nostore_sha, random_id, check_content):
824
"""See VersionedFile.add_lines."""
825
self._check_add(version_id, lines, random_id, check_content)
826
self._check_versions_present(parents)
827
return self._add(version_id, lines[:], parents, self.delta,
828
parent_texts, left_matching_blocks, nostore_sha, random_id)
830
def _check_add(self, version_id, lines, random_id, check_content):
831
"""check that version_id and lines are safe to add."""
832
if contains_whitespace(version_id):
833
raise InvalidRevisionId(version_id, self.filename)
834
self.check_not_reserved_id(version_id)
835
# Technically this could be avoided if we are happy to allow duplicate
836
# id insertion when other things than bzr core insert texts, but it
837
# seems useful for folk using the knit api directly to have some safety
838
# blanket that we can disable.
839
if not random_id and self.has_version(version_id):
840
raise RevisionAlreadyPresent(version_id, self.filename)
842
self._check_lines_not_unicode(lines)
843
self._check_lines_are_lines(lines)
845
def _add(self, version_id, lines, parents, delta, parent_texts,
846
left_matching_blocks, nostore_sha, random_id):
847
"""Add a set of lines on top of version specified by parents.
849
If delta is true, compress the text as a line-delta against
852
Any versions not present will be converted into ghosts.
854
# first thing, if the content is something we don't need to store, find
856
line_bytes = ''.join(lines)
857
digest = sha_string(line_bytes)
858
if nostore_sha == digest:
859
raise errors.ExistingContent
862
if parent_texts is None:
864
for parent in parents:
865
if self.has_version(parent):
866
present_parents.append(parent)
868
# can only compress against the left most present parent.
870
(len(present_parents) == 0 or
871
present_parents[0] != parents[0])):
874
text_length = len(line_bytes)
877
if lines[-1][-1] != '\n':
878
# copy the contents of lines.
880
options.append('no-eol')
881
lines[-1] = lines[-1] + '\n'
884
# To speed the extract of texts the delta chain is limited
885
# to a fixed number of deltas. This should minimize both
886
# I/O and the time spend applying deltas.
887
delta = self._check_should_delta(present_parents)
889
assert isinstance(version_id, str)
890
content = self.factory.make(lines, version_id)
891
if delta or (self.factory.annotated and len(present_parents) > 0):
892
# Merge annotations from parent texts if needed.
893
delta_hunks = self._merge_annotations(content, present_parents,
894
parent_texts, delta, self.factory.annotated,
895
left_matching_blocks)
898
options.append('line-delta')
899
store_lines = self.factory.lower_line_delta(delta_hunks)
900
size, bytes = self._data._record_to_data(version_id, digest,
903
options.append('fulltext')
904
# get mixed annotation + content and feed it into the
906
store_lines = self.factory.lower_fulltext(content)
907
size, bytes = self._data._record_to_data(version_id, digest,
910
access_memo = self._data.add_raw_records([size], bytes)[0]
911
self._index.add_versions(
912
((version_id, options, access_memo, parents),),
914
return digest, text_length, content
916
def check(self, progress_bar=None):
917
"""See VersionedFile.check()."""
919
def _clone_text(self, new_version_id, old_version_id, parents):
920
"""See VersionedFile.clone_text()."""
921
# FIXME RBC 20060228 make fast by only inserting an index with null
923
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
925
def get_lines(self, version_id):
926
"""See VersionedFile.get_lines()."""
927
return self.get_line_list([version_id])[0]
929
def _get_record_map(self, version_ids):
930
"""Produce a dictionary of knit records.
932
The keys are version_ids, the values are tuples of (method, content,
934
method is the way the content should be applied.
935
content is a KnitContent object.
936
digest is the SHA1 digest of this version id after all steps are done
937
next is the build-parent of the version, i.e. the leftmost ancestor.
938
If the method is fulltext, next will be None.
940
position_map = self._get_components_positions(version_ids)
941
# c = component_id, m = method, i_m = index_memo, n = next
942
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
944
for component_id, content, digest in \
945
self._data.read_records_iter(records):
946
method, index_memo, next = position_map[component_id]
947
record_map[component_id] = method, content, digest, next
951
def get_text(self, version_id):
952
"""See VersionedFile.get_text"""
953
return self.get_texts([version_id])[0]
955
def get_texts(self, version_ids):
956
return [''.join(l) for l in self.get_line_list(version_ids)]
958
def get_line_list(self, version_ids):
959
"""Return the texts of listed versions as a list of strings."""
960
for version_id in version_ids:
961
self.check_not_reserved_id(version_id)
962
text_map, content_map = self._get_content_maps(version_ids)
963
return [text_map[v] for v in version_ids]
965
_get_lf_split_line_list = get_line_list
967
def _get_content_maps(self, version_ids):
968
"""Produce maps of text and KnitContents
970
:return: (text_map, content_map) where text_map contains the texts for
971
the requested versions and content_map contains the KnitContents.
972
Both dicts take version_ids as their keys.
974
for version_id in version_ids:
975
if not self.has_version(version_id):
976
raise RevisionNotPresent(version_id, self.filename)
977
record_map = self._get_record_map(version_ids)
982
for version_id in version_ids:
985
while cursor is not None:
986
method, data, digest, next = record_map[cursor]
987
components.append((cursor, method, data, digest))
988
if cursor in content_map:
993
for component_id, method, data, digest in reversed(components):
994
if component_id in content_map:
995
content = content_map[component_id]
997
if method == 'fulltext':
998
assert content is None
999
content = self.factory.parse_fulltext(data, version_id)
1000
elif method == 'line-delta':
1001
delta = self.factory.parse_line_delta(data, version_id)
1002
content = content.copy()
1003
content._lines = self._apply_delta(content._lines,
1005
content_map[component_id] = content
1007
if 'no-eol' in self._index.get_options(version_id):
1008
content = content.copy()
1009
content.strip_last_line_newline()
1010
final_content[version_id] = content
1012
# digest here is the digest from the last applied component.
1013
text = content.text()
1014
if sha_strings(text) != digest:
1015
raise KnitCorrupt(self.filename,
1016
'sha-1 does not match %s' % version_id)
1018
text_map[version_id] = text
1019
return text_map, final_content
1022
def _apply_delta(lines, delta):
1023
"""Apply delta to lines."""
1026
for start, end, count, delta_lines in delta:
1027
lines[offset+start:offset+end] = delta_lines
1028
offset = offset + (start - end) + count
1031
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1033
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1034
if version_ids is None:
1035
version_ids = self.versions()
1037
pb = progress.DummyProgress()
1038
# we don't care about inclusions, the caller cares.
1039
# but we need to setup a list of records to visit.
1040
# we need version_id, position, length
1041
version_id_records = []
1042
requested_versions = set(version_ids)
1043
# filter for available versions
1044
for version_id in requested_versions:
1045
if not self.has_version(version_id):
1046
raise RevisionNotPresent(version_id, self.filename)
1047
# get a in-component-order queue:
1048
for version_id in self.versions():
1049
if version_id in requested_versions:
1050
index_memo = self._index.get_position(version_id)
1051
version_id_records.append((version_id, index_memo))
1053
total = len(version_id_records)
1054
for version_idx, (version_id, data, sha_value) in \
1055
enumerate(self._data.read_records_iter(version_id_records)):
1056
pb.update('Walking content.', version_idx, total)
1057
method = self._index.get_method(version_id)
1059
assert method in ('fulltext', 'line-delta')
1060
if method == 'fulltext':
1061
line_iterator = self.factory.get_fulltext_content(data)
1063
line_iterator = self.factory.get_linedelta_content(data)
1064
for line in line_iterator:
1067
pb.update('Walking content.', total, total)
1069
def iter_parents(self, version_ids):
1070
"""Iterate through the parents for many version ids.
1072
:param version_ids: An iterable yielding version_ids.
1073
:return: An iterator that yields (version_id, parents). Requested
1074
version_ids not present in the versioned file are simply skipped.
1075
The order is undefined, allowing for different optimisations in
1076
the underlying implementation.
1078
return self._index.iter_parents(version_ids)
1080
def num_versions(self):
1081
"""See VersionedFile.num_versions()."""
1082
return self._index.num_versions()
1084
__len__ = num_versions
1086
def annotate_iter(self, version_id):
1087
"""See VersionedFile.annotate_iter."""
1088
return self.factory.annotate_iter(self, version_id)
1090
def get_parents(self, version_id):
1091
"""See VersionedFile.get_parents."""
1094
# 52554 calls in 1264 872 internal down from 3674
1096
return self._index.get_parents(version_id)
1098
raise RevisionNotPresent(version_id, self.filename)
1100
def get_parents_with_ghosts(self, version_id):
1101
"""See VersionedFile.get_parents."""
1103
return self._index.get_parents_with_ghosts(version_id)
1105
raise RevisionNotPresent(version_id, self.filename)
1107
def get_ancestry(self, versions, topo_sorted=True):
1108
"""See VersionedFile.get_ancestry."""
1109
if isinstance(versions, basestring):
1110
versions = [versions]
1113
return self._index.get_ancestry(versions, topo_sorted)
1115
def get_ancestry_with_ghosts(self, versions):
1116
"""See VersionedFile.get_ancestry_with_ghosts."""
1117
if isinstance(versions, basestring):
1118
versions = [versions]
1121
return self._index.get_ancestry_with_ghosts(versions)
1123
def plan_merge(self, ver_a, ver_b):
1124
"""See VersionedFile.plan_merge."""
1125
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1126
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1127
annotated_a = self.annotate(ver_a)
1128
annotated_b = self.annotate(ver_b)
1129
return merge._plan_annotate_merge(annotated_a, annotated_b,
1130
ancestors_a, ancestors_b)
1133
class _KnitComponentFile(object):
1134
"""One of the files used to implement a knit database"""
1136
def __init__(self, transport, filename, mode, file_mode=None,
1137
create_parent_dir=False, dir_mode=None):
1138
self._transport = transport
1139
self._filename = filename
1141
self._file_mode = file_mode
1142
self._dir_mode = dir_mode
1143
self._create_parent_dir = create_parent_dir
1144
self._need_to_create = False
1146
def _full_path(self):
1147
"""Return the full path to this file."""
1148
return self._transport.base + self._filename
1150
def check_header(self, fp):
1151
line = fp.readline()
1153
# An empty file can actually be treated as though the file doesn't
1155
raise errors.NoSuchFile(self._full_path())
1156
if line != self.HEADER:
1157
raise KnitHeaderError(badline=line,
1158
filename=self._transport.abspath(self._filename))
1161
return '%s(%s)' % (self.__class__.__name__, self._filename)
1164
class _KnitIndex(_KnitComponentFile):
1165
"""Manages knit index file.
1167
The index is already kept in memory and read on startup, to enable
1168
fast lookups of revision information. The cursor of the index
1169
file is always pointing to the end, making it easy to append
1172
_cache is a cache for fast mapping from version id to a Index
1175
_history is a cache for fast mapping from indexes to version ids.
1177
The index data format is dictionary compressed when it comes to
1178
parent references; a index entry may only have parents that with a
1179
lover index number. As a result, the index is topological sorted.
1181
Duplicate entries may be written to the index for a single version id
1182
if this is done then the latter one completely replaces the former:
1183
this allows updates to correct version and parent information.
1184
Note that the two entries may share the delta, and that successive
1185
annotations and references MUST point to the first entry.
1187
The index file on disc contains a header, followed by one line per knit
1188
record. The same revision can be present in an index file more than once.
1189
The first occurrence gets assigned a sequence number starting from 0.
1191
The format of a single line is
1192
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1193
REVISION_ID is a utf8-encoded revision id
1194
FLAGS is a comma separated list of flags about the record. Values include
1195
no-eol, line-delta, fulltext.
1196
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1197
that the the compressed data starts at.
1198
LENGTH is the ascii representation of the length of the data file.
1199
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1201
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1202
revision id already in the knit that is a parent of REVISION_ID.
1203
The ' :' marker is the end of record marker.
1206
when a write is interrupted to the index file, it will result in a line
1207
that does not end in ' :'. If the ' :' is not present at the end of a line,
1208
or at the end of the file, then the record that is missing it will be
1209
ignored by the parser.
1211
When writing new records to the index file, the data is preceded by '\n'
1212
to ensure that records always start on new lines even if the last write was
1213
interrupted. As a result its normal for the last line in the index to be
1214
missing a trailing newline. One can be added with no harmful effects.
1217
HEADER = "# bzr knit index 8\n"
1219
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1220
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1222
def _cache_version(self, version_id, options, pos, size, parents):
1223
"""Cache a version record in the history array and index cache.
1225
This is inlined into _load_data for performance. KEEP IN SYNC.
1226
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1229
# only want the _history index to reference the 1st index entry
1231
if version_id not in self._cache:
1232
index = len(self._history)
1233
self._history.append(version_id)
1235
index = self._cache[version_id][5]
1236
self._cache[version_id] = (version_id,
1243
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1244
create_parent_dir=False, delay_create=False, dir_mode=None):
1245
_KnitComponentFile.__init__(self, transport, filename, mode,
1246
file_mode=file_mode,
1247
create_parent_dir=create_parent_dir,
1250
# position in _history is the 'official' index for a revision
1251
# but the values may have come from a newer entry.
1252
# so - wc -l of a knit index is != the number of unique names
1256
fp = self._transport.get(self._filename)
1258
# _load_data may raise NoSuchFile if the target knit is
1260
_load_data(self, fp)
1264
if mode != 'w' or not create:
1267
self._need_to_create = True
1269
self._transport.put_bytes_non_atomic(
1270
self._filename, self.HEADER, mode=self._file_mode)
1272
def get_graph(self):
1273
"""Return a list of the node:parents lists from this knit index."""
1274
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1276
def get_ancestry(self, versions, topo_sorted=True):
1277
"""See VersionedFile.get_ancestry."""
1278
# get a graph of all the mentioned versions:
1280
pending = set(versions)
1283
version = pending.pop()
1286
parents = [p for p in cache[version][4] if p in cache]
1288
raise RevisionNotPresent(version, self._filename)
1289
# if not completed and not a ghost
1290
pending.update([p for p in parents if p not in graph])
1291
graph[version] = parents
1294
return topo_sort(graph.items())
1296
def get_ancestry_with_ghosts(self, versions):
1297
"""See VersionedFile.get_ancestry_with_ghosts."""
1298
# get a graph of all the mentioned versions:
1299
self.check_versions_present(versions)
1302
pending = set(versions)
1304
version = pending.pop()
1306
parents = cache[version][4]
1312
pending.update([p for p in parents if p not in graph])
1313
graph[version] = parents
1314
return topo_sort(graph.items())
1316
def iter_parents(self, version_ids):
1317
"""Iterate through the parents for many version ids.
1319
:param version_ids: An iterable yielding version_ids.
1320
:return: An iterator that yields (version_id, parents). Requested
1321
version_ids not present in the versioned file are simply skipped.
1322
The order is undefined, allowing for different optimisations in
1323
the underlying implementation.
1325
for version_id in version_ids:
1327
yield version_id, tuple(self.get_parents(version_id))
1331
def num_versions(self):
1332
return len(self._history)
1334
__len__ = num_versions
1336
def get_versions(self):
1337
"""Get all the versions in the file. not topologically sorted."""
1338
return self._history
1340
def _version_list_to_index(self, versions):
1343
for version in versions:
1344
if version in cache:
1345
# -- inlined lookup() --
1346
result_list.append(str(cache[version][5]))
1347
# -- end lookup () --
1349
result_list.append('.' + version)
1350
return ' '.join(result_list)
1352
def add_version(self, version_id, options, index_memo, parents):
1353
"""Add a version record to the index."""
1354
self.add_versions(((version_id, options, index_memo, parents),))
1356
def add_versions(self, versions, random_id=False):
1357
"""Add multiple versions to the index.
1359
:param versions: a list of tuples:
1360
(version_id, options, pos, size, parents).
1361
:param random_id: If True the ids being added were randomly generated
1362
and no check for existence will be performed.
1365
orig_history = self._history[:]
1366
orig_cache = self._cache.copy()
1369
for version_id, options, (index, pos, size), parents in versions:
1370
line = "\n%s %s %s %s %s :" % (version_id,
1374
self._version_list_to_index(parents))
1375
assert isinstance(line, str), \
1376
'content must be utf-8 encoded: %r' % (line,)
1378
self._cache_version(version_id, options, pos, size, parents)
1379
if not self._need_to_create:
1380
self._transport.append_bytes(self._filename, ''.join(lines))
1383
sio.write(self.HEADER)
1384
sio.writelines(lines)
1386
self._transport.put_file_non_atomic(self._filename, sio,
1387
create_parent_dir=self._create_parent_dir,
1388
mode=self._file_mode,
1389
dir_mode=self._dir_mode)
1390
self._need_to_create = False
1392
# If any problems happen, restore the original values and re-raise
1393
self._history = orig_history
1394
self._cache = orig_cache
1397
def has_version(self, version_id):
1398
"""True if the version is in the index."""
1399
return version_id in self._cache
1401
def get_position(self, version_id):
1402
"""Return details needed to access the version.
1404
.kndx indices do not support split-out data, so return None for the
1407
:return: a tuple (None, data position, size) to hand to the access
1408
logic to get the record.
1410
entry = self._cache[version_id]
1411
return None, entry[2], entry[3]
1413
def get_method(self, version_id):
1414
"""Return compression method of specified version."""
1415
options = self._cache[version_id][1]
1416
if 'fulltext' in options:
1419
if 'line-delta' not in options:
1420
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1423
def get_options(self, version_id):
1424
"""Return a string represention options.
1428
return self._cache[version_id][1]
1430
def get_parents(self, version_id):
1431
"""Return parents of specified version ignoring ghosts."""
1432
return [parent for parent in self._cache[version_id][4]
1433
if parent in self._cache]
1435
def get_parents_with_ghosts(self, version_id):
1436
"""Return parents of specified version with ghosts."""
1437
return self._cache[version_id][4]
1439
def check_versions_present(self, version_ids):
1440
"""Check that all specified versions are present."""
1442
for version_id in version_ids:
1443
if version_id not in cache:
1444
raise RevisionNotPresent(version_id, self._filename)
1447
class KnitGraphIndex(object):
1448
"""A knit index that builds on GraphIndex."""
1450
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1451
"""Construct a KnitGraphIndex on a graph_index.
1453
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1454
:param deltas: Allow delta-compressed records.
1455
:param add_callback: If not None, allow additions to the index and call
1456
this callback with a list of added GraphIndex nodes:
1457
[(node, value, node_refs), ...]
1458
:param parents: If True, record knits parents, if not do not record
1461
self._graph_index = graph_index
1462
self._deltas = deltas
1463
self._add_callback = add_callback
1464
self._parents = parents
1465
if deltas and not parents:
1466
raise KnitCorrupt(self, "Cannot do delta compression without "
1469
def _get_entries(self, keys, check_present=False):
1470
"""Get the entries for keys.
1472
:param keys: An iterable of index keys, - 1-tuples.
1477
for node in self._graph_index.iter_entries(keys):
1479
found_keys.add(node[1])
1481
# adapt parentless index to the rest of the code.
1482
for node in self._graph_index.iter_entries(keys):
1483
yield node[0], node[1], node[2], ()
1484
found_keys.add(node[1])
1486
missing_keys = keys.difference(found_keys)
1488
raise RevisionNotPresent(missing_keys.pop(), self)
1490
def _present_keys(self, version_ids):
1492
node[1] for node in self._get_entries(version_ids)])
1494
def _parentless_ancestry(self, versions):
1495
"""Honour the get_ancestry API for parentless knit indices."""
1496
wanted_keys = self._version_ids_to_keys(versions)
1497
present_keys = self._present_keys(wanted_keys)
1498
missing = set(wanted_keys).difference(present_keys)
1500
raise RevisionNotPresent(missing.pop(), self)
1501
return list(self._keys_to_version_ids(present_keys))
1503
def get_ancestry(self, versions, topo_sorted=True):
1504
"""See VersionedFile.get_ancestry."""
1505
if not self._parents:
1506
return self._parentless_ancestry(versions)
1507
# XXX: This will do len(history) index calls - perhaps
1508
# it should be altered to be a index core feature?
1509
# get a graph of all the mentioned versions:
1512
versions = self._version_ids_to_keys(versions)
1513
pending = set(versions)
1515
# get all pending nodes
1516
this_iteration = pending
1517
new_nodes = self._get_entries(this_iteration)
1520
for (index, key, value, node_refs) in new_nodes:
1521
# dont ask for ghosties - otherwise
1522
# we we can end up looping with pending
1523
# being entirely ghosted.
1524
graph[key] = [parent for parent in node_refs[0]
1525
if parent not in ghosts]
1527
for parent in graph[key]:
1528
# dont examine known nodes again
1533
ghosts.update(this_iteration.difference(found))
1534
if versions.difference(graph):
1535
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1537
result_keys = topo_sort(graph.items())
1539
result_keys = graph.iterkeys()
1540
return [key[0] for key in result_keys]
1542
def get_ancestry_with_ghosts(self, versions):
1543
"""See VersionedFile.get_ancestry."""
1544
if not self._parents:
1545
return self._parentless_ancestry(versions)
1546
# XXX: This will do len(history) index calls - perhaps
1547
# it should be altered to be a index core feature?
1548
# get a graph of all the mentioned versions:
1550
versions = self._version_ids_to_keys(versions)
1551
pending = set(versions)
1553
# get all pending nodes
1554
this_iteration = pending
1555
new_nodes = self._get_entries(this_iteration)
1557
for (index, key, value, node_refs) in new_nodes:
1558
graph[key] = node_refs[0]
1560
for parent in graph[key]:
1561
# dont examine known nodes again
1565
missing_versions = this_iteration.difference(graph)
1566
missing_needed = versions.intersection(missing_versions)
1568
raise RevisionNotPresent(missing_needed.pop(), self)
1569
for missing_version in missing_versions:
1570
# add a key, no parents
1571
graph[missing_version] = []
1572
pending.discard(missing_version) # don't look for it
1573
result_keys = topo_sort(graph.items())
1574
return [key[0] for key in result_keys]
1576
def get_graph(self):
1577
"""Return a list of the node:parents lists from this knit index."""
1578
if not self._parents:
1579
return [(key, ()) for key in self.get_versions()]
1581
for index, key, value, refs in self._graph_index.iter_all_entries():
1582
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1585
def iter_parents(self, version_ids):
1586
"""Iterate through the parents for many version ids.
1588
:param version_ids: An iterable yielding version_ids.
1589
:return: An iterator that yields (version_id, parents). Requested
1590
version_ids not present in the versioned file are simply skipped.
1591
The order is undefined, allowing for different optimisations in
1592
the underlying implementation.
1595
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1597
present_parents = set()
1598
for node in all_nodes:
1599
all_parents.update(node[3][0])
1600
# any node we are querying must be present
1601
present_parents.add(node[1])
1602
unknown_parents = all_parents.difference(present_parents)
1603
present_parents.update(self._present_keys(unknown_parents))
1604
for node in all_nodes:
1606
for parent in node[3][0]:
1607
if parent in present_parents:
1608
parents.append(parent[0])
1609
yield node[1][0], tuple(parents)
1611
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1612
yield node[1][0], ()
1614
def num_versions(self):
1615
return len(list(self._graph_index.iter_all_entries()))
1617
__len__ = num_versions
1619
def get_versions(self):
1620
"""Get all the versions in the file. not topologically sorted."""
1621
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1623
def has_version(self, version_id):
1624
"""True if the version is in the index."""
1625
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1627
def _keys_to_version_ids(self, keys):
1628
return tuple(key[0] for key in keys)
1630
def get_position(self, version_id):
1631
"""Return details needed to access the version.
1633
:return: a tuple (index, data position, size) to hand to the access
1634
logic to get the record.
1636
node = self._get_node(version_id)
1637
bits = node[2][1:].split(' ')
1638
return node[0], int(bits[0]), int(bits[1])
1640
def get_method(self, version_id):
1641
"""Return compression method of specified version."""
1642
if not self._deltas:
1644
return self._parent_compression(self._get_node(version_id)[3][1])
1646
def _parent_compression(self, reference_list):
1647
# use the second reference list to decide if this is delta'd or not.
1648
if len(reference_list):
1653
def _get_node(self, version_id):
1654
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1656
def get_options(self, version_id):
1657
"""Return a string represention options.
1661
node = self._get_node(version_id)
1662
if not self._deltas:
1663
options = ['fulltext']
1665
options = [self._parent_compression(node[3][1])]
1666
if node[2][0] == 'N':
1667
options.append('no-eol')
1670
def get_parents(self, version_id):
1671
"""Return parents of specified version ignoring ghosts."""
1672
parents = list(self.iter_parents([version_id]))
1675
raise errors.RevisionNotPresent(version_id, self)
1676
return parents[0][1]
1678
def get_parents_with_ghosts(self, version_id):
1679
"""Return parents of specified version with ghosts."""
1680
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1681
check_present=True))
1682
if not self._parents:
1684
return self._keys_to_version_ids(nodes[0][3][0])
1686
def check_versions_present(self, version_ids):
1687
"""Check that all specified versions are present."""
1688
keys = self._version_ids_to_keys(version_ids)
1689
present = self._present_keys(keys)
1690
missing = keys.difference(present)
1692
raise RevisionNotPresent(missing.pop(), self)
1694
def add_version(self, version_id, options, access_memo, parents):
1695
"""Add a version record to the index."""
1696
return self.add_versions(((version_id, options, access_memo, parents),))
1698
def add_versions(self, versions, random_id=False):
1699
"""Add multiple versions to the index.
1701
This function does not insert data into the Immutable GraphIndex
1702
backing the KnitGraphIndex, instead it prepares data for insertion by
1703
the caller and checks that it is safe to insert then calls
1704
self._add_callback with the prepared GraphIndex nodes.
1706
:param versions: a list of tuples:
1707
(version_id, options, pos, size, parents).
1708
:param random_id: If True the ids being added were randomly generated
1709
and no check for existence will be performed.
1711
if not self._add_callback:
1712
raise errors.ReadOnlyError(self)
1713
# we hope there are no repositories with inconsistent parentage
1718
for (version_id, options, access_memo, parents) in versions:
1719
index, pos, size = access_memo
1720
key = (version_id, )
1721
parents = tuple((parent, ) for parent in parents)
1722
if 'no-eol' in options:
1726
value += "%d %d" % (pos, size)
1727
if not self._deltas:
1728
if 'line-delta' in options:
1729
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1732
if 'line-delta' in options:
1733
node_refs = (parents, (parents[0],))
1735
node_refs = (parents, ())
1737
node_refs = (parents, )
1740
raise KnitCorrupt(self, "attempt to add node with parents "
1741
"in parentless index.")
1743
keys[key] = (value, node_refs)
1745
present_nodes = self._get_entries(keys)
1746
for (index, key, value, node_refs) in present_nodes:
1747
if (value, node_refs) != keys[key]:
1748
raise KnitCorrupt(self, "inconsistent details in add_versions"
1749
": %s %s" % ((value, node_refs), keys[key]))
1753
for key, (value, node_refs) in keys.iteritems():
1754
result.append((key, value, node_refs))
1756
for key, (value, node_refs) in keys.iteritems():
1757
result.append((key, value))
1758
self._add_callback(result)
1760
def _version_ids_to_keys(self, version_ids):
1761
return set((version_id, ) for version_id in version_ids)
1764
class _KnitAccess(object):
1765
"""Access to knit records in a .knit file."""
1767
def __init__(self, transport, filename, _file_mode, _dir_mode,
1768
_need_to_create, _create_parent_dir):
1769
"""Create a _KnitAccess for accessing and inserting data.
1771
:param transport: The transport the .knit is located on.
1772
:param filename: The filename of the .knit.
1774
self._transport = transport
1775
self._filename = filename
1776
self._file_mode = _file_mode
1777
self._dir_mode = _dir_mode
1778
self._need_to_create = _need_to_create
1779
self._create_parent_dir = _create_parent_dir
1781
def add_raw_records(self, sizes, raw_data):
1782
"""Add raw knit bytes to a storage area.
1784
The data is spooled to whereever the access method is storing data.
1786
:param sizes: An iterable containing the size of each raw data segment.
1787
:param raw_data: A bytestring containing the data.
1788
:return: A list of memos to retrieve the record later. Each memo is a
1789
tuple - (index, pos, length), where the index field is always None
1790
for the .knit access method.
1792
assert type(raw_data) == str, \
1793
'data must be plain bytes was %s' % type(raw_data)
1794
if not self._need_to_create:
1795
base = self._transport.append_bytes(self._filename, raw_data)
1797
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1798
create_parent_dir=self._create_parent_dir,
1799
mode=self._file_mode,
1800
dir_mode=self._dir_mode)
1801
self._need_to_create = False
1805
result.append((None, base, size))
1810
"""IFF this data access has its own storage area, initialise it.
1814
self._transport.put_bytes_non_atomic(self._filename, '',
1815
mode=self._file_mode)
1817
def open_file(self):
1818
"""IFF this data access can be represented as a single file, open it.
1820
For knits that are not mapped to a single file on disk this will
1823
:return: None or a file handle.
1826
return self._transport.get(self._filename)
1831
def get_raw_records(self, memos_for_retrieval):
1832
"""Get the raw bytes for a records.
1834
:param memos_for_retrieval: An iterable containing the (index, pos,
1835
length) memo for retrieving the bytes. The .knit method ignores
1836
the index as there is always only a single file.
1837
:return: An iterator over the bytes of the records.
1839
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1840
for pos, data in self._transport.readv(self._filename, read_vector):
1844
class _PackAccess(object):
1845
"""Access to knit records via a collection of packs."""
1847
def __init__(self, index_to_packs, writer=None):
1848
"""Create a _PackAccess object.
1850
:param index_to_packs: A dict mapping index objects to the transport
1851
and file names for obtaining data.
1852
:param writer: A tuple (pack.ContainerWriter, write_index) which
1853
contains the pack to write, and the index that reads from it will
1857
self.container_writer = writer[0]
1858
self.write_index = writer[1]
1860
self.container_writer = None
1861
self.write_index = None
1862
self.indices = index_to_packs
1864
def add_raw_records(self, sizes, raw_data):
1865
"""Add raw knit bytes to a storage area.
1867
The data is spooled to the container writer in one bytes-record per
1870
:param sizes: An iterable containing the size of each raw data segment.
1871
:param raw_data: A bytestring containing the data.
1872
:return: A list of memos to retrieve the record later. Each memo is a
1873
tuple - (index, pos, length), where the index field is the
1874
write_index object supplied to the PackAccess object.
1876
assert type(raw_data) == str, \
1877
'data must be plain bytes was %s' % type(raw_data)
1881
p_offset, p_length = self.container_writer.add_bytes_record(
1882
raw_data[offset:offset+size], [])
1884
result.append((self.write_index, p_offset, p_length))
1888
"""Pack based knits do not get individually created."""
1890
def get_raw_records(self, memos_for_retrieval):
1891
"""Get the raw bytes for a records.
1893
:param memos_for_retrieval: An iterable containing the (index, pos,
1894
length) memo for retrieving the bytes. The Pack access method
1895
looks up the pack to use for a given record in its index_to_pack
1897
:return: An iterator over the bytes of the records.
1899
# first pass, group into same-index requests
1901
current_index = None
1902
for (index, offset, length) in memos_for_retrieval:
1903
if current_index == index:
1904
current_list.append((offset, length))
1906
if current_index is not None:
1907
request_lists.append((current_index, current_list))
1908
current_index = index
1909
current_list = [(offset, length)]
1910
# handle the last entry
1911
if current_index is not None:
1912
request_lists.append((current_index, current_list))
1913
for index, offsets in request_lists:
1914
transport, path = self.indices[index]
1915
reader = pack.make_readv_reader(transport, path, offsets)
1916
for names, read_func in reader.iter_records():
1917
yield read_func(None)
1919
def open_file(self):
1920
"""Pack based knits have no single file."""
1923
def set_writer(self, writer, index, (transport, packname)):
1924
"""Set a writer to use for adding data."""
1925
self.indices[index] = (transport, packname)
1926
self.container_writer = writer
1927
self.write_index = index
1930
class _KnitData(object):
1931
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1933
The KnitData class provides the logic for parsing and using knit records,
1934
making use of an access method for the low level read and write operations.
1937
def __init__(self, access):
1938
"""Create a KnitData object.
1940
:param access: The access method to use. Access methods such as
1941
_KnitAccess manage the insertion of raw records and the subsequent
1942
retrieval of the same.
1944
self._access = access
1945
self._checked = False
1946
# TODO: jam 20060713 conceptually, this could spill to disk
1947
# if the cached size gets larger than a certain amount
1948
# but it complicates the model a bit, so for now just use
1949
# a simple dictionary
1951
self._do_cache = False
1953
def enable_cache(self):
1954
"""Enable caching of reads."""
1955
self._do_cache = True
1957
def clear_cache(self):
1958
"""Clear the record cache."""
1959
self._do_cache = False
1962
def _open_file(self):
1963
return self._access.open_file()
1965
def _record_to_data(self, version_id, digest, lines):
1966
"""Convert version_id, digest, lines into a raw data block.
1968
:return: (len, a StringIO instance with the raw data ready to read.)
1970
bytes = (''.join(chain(
1971
["version %s %d %s\n" % (version_id,
1975
["end %s\n" % version_id])))
1976
assert bytes.__class__ == str
1977
compressed_bytes = bytes_to_gzip(bytes)
1978
return len(compressed_bytes), compressed_bytes
1980
def add_raw_records(self, sizes, raw_data):
1981
"""Append a prepared record to the data file.
1983
:param sizes: An iterable containing the size of each raw data segment.
1984
:param raw_data: A bytestring containing the data.
1985
:return: a list of index data for the way the data was stored.
1986
See the access method add_raw_records documentation for more
1989
return self._access.add_raw_records(sizes, raw_data)
1991
def _parse_record_header(self, version_id, raw_data):
1992
"""Parse a record header for consistency.
1994
:return: the header and the decompressor stream.
1995
as (stream, header_record)
1997
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
1999
rec = self._check_header(version_id, df.readline())
2000
except Exception, e:
2001
raise KnitCorrupt(self._access,
2002
"While reading {%s} got %s(%s)"
2003
% (version_id, e.__class__.__name__, str(e)))
2006
def _check_header(self, version_id, line):
2009
raise KnitCorrupt(self._access,
2010
'unexpected number of elements in record header')
2011
if rec[1] != version_id:
2012
raise KnitCorrupt(self._access,
2013
'unexpected version, wanted %r, got %r'
2014
% (version_id, rec[1]))
2017
def _parse_record(self, version_id, data):
2019
# 4168 calls in 2880 217 internal
2020
# 4168 calls to _parse_record_header in 2121
2021
# 4168 calls to readlines in 330
2022
df = GzipFile(mode='rb', fileobj=StringIO(data))
2025
record_contents = df.readlines()
2026
except Exception, e:
2027
raise KnitCorrupt(self._access,
2028
"While reading {%s} got %s(%s)"
2029
% (version_id, e.__class__.__name__, str(e)))
2030
header = record_contents.pop(0)
2031
rec = self._check_header(version_id, header)
2033
last_line = record_contents.pop()
2034
if len(record_contents) != int(rec[2]):
2035
raise KnitCorrupt(self._access,
2036
'incorrect number of lines %s != %s'
2038
% (len(record_contents), int(rec[2]),
2040
if last_line != 'end %s\n' % rec[1]:
2041
raise KnitCorrupt(self._access,
2042
'unexpected version end line %r, wanted %r'
2043
% (last_line, version_id))
2045
return record_contents, rec[3]
2047
def read_records_iter_raw(self, records):
2048
"""Read text records from data file and yield raw data.
2050
This unpacks enough of the text record to validate the id is
2051
as expected but thats all.
2053
# setup an iterator of the external records:
2054
# uses readv so nice and fast we hope.
2056
# grab the disk data needed.
2058
# Don't check _cache if it is empty
2059
needed_offsets = [index_memo for version_id, index_memo
2061
if version_id not in self._cache]
2063
needed_offsets = [index_memo for version_id, index_memo
2066
raw_records = self._access.get_raw_records(needed_offsets)
2068
for version_id, index_memo in records:
2069
if version_id in self._cache:
2070
# This data has already been validated
2071
data = self._cache[version_id]
2073
data = raw_records.next()
2075
self._cache[version_id] = data
2077
# validate the header
2078
df, rec = self._parse_record_header(version_id, data)
2080
yield version_id, data
2082
def read_records_iter(self, records):
2083
"""Read text records from data file and yield result.
2085
The result will be returned in whatever is the fastest to read.
2086
Not by the order requested. Also, multiple requests for the same
2087
record will only yield 1 response.
2088
:param records: A list of (version_id, pos, len) entries
2089
:return: Yields (version_id, contents, digest) in the order
2090
read, not the order requested
2096
# Skip records we have alread seen
2097
yielded_records = set()
2098
needed_records = set()
2099
for record in records:
2100
if record[0] in self._cache:
2101
if record[0] in yielded_records:
2103
yielded_records.add(record[0])
2104
data = self._cache[record[0]]
2105
content, digest = self._parse_record(record[0], data)
2106
yield (record[0], content, digest)
2108
needed_records.add(record)
2109
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2111
needed_records = sorted(set(records), key=operator.itemgetter(1))
2113
if not needed_records:
2116
# The transport optimizes the fetching as well
2117
# (ie, reads continuous ranges.)
2118
raw_data = self._access.get_raw_records(
2119
[index_memo for version_id, index_memo in needed_records])
2121
for (version_id, index_memo), data in \
2122
izip(iter(needed_records), raw_data):
2123
content, digest = self._parse_record(version_id, data)
2125
self._cache[version_id] = data
2126
yield version_id, content, digest
2128
def read_records(self, records):
2129
"""Read records into a dictionary."""
2131
for record_id, content, digest in \
2132
self.read_records_iter(records):
2133
components[record_id] = (content, digest)
2137
class InterKnit(InterVersionedFile):
2138
"""Optimised code paths for knit to knit operations."""
2140
_matching_file_from_factory = KnitVersionedFile
2141
_matching_file_to_factory = KnitVersionedFile
2144
def is_compatible(source, target):
2145
"""Be compatible with knits. """
2147
return (isinstance(source, KnitVersionedFile) and
2148
isinstance(target, KnitVersionedFile))
2149
except AttributeError:
2152
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2153
"""See InterVersionedFile.join."""
2154
assert isinstance(self.source, KnitVersionedFile)
2155
assert isinstance(self.target, KnitVersionedFile)
2157
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2162
pb = ui.ui_factory.nested_progress_bar()
2164
version_ids = list(version_ids)
2165
if None in version_ids:
2166
version_ids.remove(None)
2168
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2169
this_versions = set(self.target._index.get_versions())
2170
# XXX: For efficiency we should not look at the whole index,
2171
# we only need to consider the referenced revisions - they
2172
# must all be present, or the method must be full-text.
2173
# TODO, RBC 20070919
2174
needed_versions = self.source_ancestry - this_versions
2176
if not needed_versions:
2178
full_list = topo_sort(self.source.get_graph())
2180
version_list = [i for i in full_list if (not self.target.has_version(i)
2181
and i in needed_versions)]
2185
copy_queue_records = []
2187
for version_id in version_list:
2188
options = self.source._index.get_options(version_id)
2189
parents = self.source._index.get_parents_with_ghosts(version_id)
2190
# check that its will be a consistent copy:
2191
for parent in parents:
2192
# if source has the parent, we must :
2193
# * already have it or
2194
# * have it scheduled already
2195
# otherwise we don't care
2196
assert (self.target.has_version(parent) or
2197
parent in copy_set or
2198
not self.source.has_version(parent))
2199
index_memo = self.source._index.get_position(version_id)
2200
copy_queue_records.append((version_id, index_memo))
2201
copy_queue.append((version_id, options, parents))
2202
copy_set.add(version_id)
2204
# data suck the join:
2206
total = len(version_list)
2209
for (version_id, raw_data), \
2210
(version_id2, options, parents) in \
2211
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2213
assert version_id == version_id2, 'logic error, inconsistent results'
2215
pb.update("Joining knit", count, total)
2216
raw_records.append((version_id, options, parents, len(raw_data)))
2217
raw_datum.append(raw_data)
2218
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2224
InterVersionedFile.register_optimiser(InterKnit)
2227
class WeaveToKnit(InterVersionedFile):
2228
"""Optimised code paths for weave to knit operations."""
2230
_matching_file_from_factory = bzrlib.weave.WeaveFile
2231
_matching_file_to_factory = KnitVersionedFile
2234
def is_compatible(source, target):
2235
"""Be compatible with weaves to knits."""
2237
return (isinstance(source, bzrlib.weave.Weave) and
2238
isinstance(target, KnitVersionedFile))
2239
except AttributeError:
2242
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2243
"""See InterVersionedFile.join."""
2244
assert isinstance(self.source, bzrlib.weave.Weave)
2245
assert isinstance(self.target, KnitVersionedFile)
2247
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2252
pb = ui.ui_factory.nested_progress_bar()
2254
version_ids = list(version_ids)
2256
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2257
this_versions = set(self.target._index.get_versions())
2258
needed_versions = self.source_ancestry - this_versions
2260
if not needed_versions:
2262
full_list = topo_sort(self.source.get_graph())
2264
version_list = [i for i in full_list if (not self.target.has_version(i)
2265
and i in needed_versions)]
2269
total = len(version_list)
2270
for version_id in version_list:
2271
pb.update("Converting to knit", count, total)
2272
parents = self.source.get_parents(version_id)
2273
# check that its will be a consistent copy:
2274
for parent in parents:
2275
# if source has the parent, we must already have it
2276
assert (self.target.has_version(parent))
2277
self.target.add_lines(
2278
version_id, parents, self.source.get_lines(version_id))
2285
InterVersionedFile.register_optimiser(WeaveToKnit)
2288
# Deprecated, use PatienceSequenceMatcher instead
2289
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2292
def annotate_knit(knit, revision_id):
2293
"""Annotate a knit with no cached annotations.
2295
This implementation is for knits with no cached annotations.
2296
It will work for knits with cached annotations, but this is not
2299
ancestry = knit.get_ancestry(revision_id)
2300
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2302
for candidate in ancestry:
2303
if candidate in annotations:
2305
parents = knit.get_parents(candidate)
2306
if len(parents) == 0:
2308
elif knit._index.get_method(candidate) != 'line-delta':
2311
parent, sha1, noeol, delta = knit.get_delta(candidate)
2312
blocks = KnitContent.get_line_delta_blocks(delta,
2313
fulltext[parents[0]], fulltext[candidate])
2314
annotations[candidate] = list(annotate.reannotate([annotations[p]
2315
for p in parents], fulltext[candidate], candidate, blocks))
2316
return iter(annotations[revision_id])
2320
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2322
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
