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, plain=False):
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)])
267
:param plain: If True, the lines are returned as a plain
268
list, not as a list of tuples, i.e.
269
(start, end, count, [1..count newline])
276
def cache_and_return(line):
277
origin, text = line.split(' ', 1)
278
return cache.setdefault(origin, origin), text
280
# walk through the lines parsing.
281
# Note that the plain test is explicitly pulled out of the
282
# loop to minimise any performance impact
285
start, end, count = [int(n) for n in header.split(',')]
286
contents = [next().split(' ', 1)[1] for i in xrange(count)]
287
result.append((start, end, count, contents))
290
start, end, count = [int(n) for n in header.split(',')]
291
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
292
result.append((start, end, count, contents))
295
def get_fulltext_content(self, lines):
296
"""Extract just the content lines from a fulltext."""
297
return (line.split(' ', 1)[1] for line in lines)
299
def get_linedelta_content(self, lines):
300
"""Extract just the content from a line delta.
302
This doesn't return all of the extra information stored in a delta.
303
Only the actual content lines.
308
header = header.split(',')
309
count = int(header[2])
310
for i in xrange(count):
311
origin, text = next().split(' ', 1)
314
def lower_fulltext(self, content):
315
"""convert a fulltext content record into a serializable form.
317
see parse_fulltext which this inverts.
319
# TODO: jam 20070209 We only do the caching thing to make sure that
320
# the origin is a valid utf-8 line, eventually we could remove it
321
return ['%s %s' % (o, t) for o, t in content._lines]
323
def lower_line_delta(self, delta):
324
"""convert a delta into a serializable form.
326
See parse_line_delta which this inverts.
328
# TODO: jam 20070209 We only do the caching thing to make sure that
329
# the origin is a valid utf-8 line, eventually we could remove it
331
for start, end, c, lines in delta:
332
out.append('%d,%d,%d\n' % (start, end, c))
333
out.extend(origin + ' ' + text
334
for origin, text in lines)
337
def annotate_iter(self, knit, version_id):
338
content = knit._get_content(version_id)
339
return content.annotate_iter()
342
class KnitPlainFactory(object):
343
"""Factory for creating plain Content objects."""
347
def make(self, lines, version_id):
348
return PlainKnitContent(lines, version_id)
350
def parse_fulltext(self, content, version_id):
351
"""This parses an unannotated fulltext.
353
Note that this is not a noop - the internal representation
354
has (versionid, line) - its just a constant versionid.
356
return self.make(content, version_id)
358
def parse_line_delta_iter(self, lines, version_id):
360
num_lines = len(lines)
361
while cur < num_lines:
364
start, end, c = [int(n) for n in header.split(',')]
365
yield start, end, c, lines[cur:cur+c]
368
def parse_line_delta(self, lines, version_id):
369
return list(self.parse_line_delta_iter(lines, version_id))
371
def get_fulltext_content(self, lines):
372
"""Extract just the content lines from a fulltext."""
375
def get_linedelta_content(self, lines):
376
"""Extract just the content from a line delta.
378
This doesn't return all of the extra information stored in a delta.
379
Only the actual content lines.
384
header = header.split(',')
385
count = int(header[2])
386
for i in xrange(count):
389
def lower_fulltext(self, content):
390
return content.text()
392
def lower_line_delta(self, delta):
394
for start, end, c, lines in delta:
395
out.append('%d,%d,%d\n' % (start, end, c))
399
def annotate_iter(self, knit, version_id):
400
return annotate_knit(knit, version_id)
403
def make_empty_knit(transport, relpath):
404
"""Construct a empty knit at the specified location."""
405
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
408
class KnitVersionedFile(VersionedFile):
409
"""Weave-like structure with faster random access.
411
A knit stores a number of texts and a summary of the relationships
412
between them. Texts are identified by a string version-id. Texts
413
are normally stored and retrieved as a series of lines, but can
414
also be passed as single strings.
416
Lines are stored with the trailing newline (if any) included, to
417
avoid special cases for files with no final newline. Lines are
418
composed of 8-bit characters, not unicode. The combination of
419
these approaches should mean any 'binary' file can be safely
420
stored and retrieved.
423
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
424
factory=None, basis_knit=DEPRECATED_PARAMETER, delta=True,
425
create=False, create_parent_dir=False, delay_create=False,
426
dir_mode=None, index=None, access_method=None):
427
"""Construct a knit at location specified by relpath.
429
:param create: If not True, only open an existing knit.
430
:param create_parent_dir: If True, create the parent directory if
431
creating the file fails. (This is used for stores with
432
hash-prefixes that may not exist yet)
433
:param delay_create: The calling code is aware that the knit won't
434
actually be created until the first data is stored.
435
:param index: An index to use for the knit.
437
if deprecated_passed(basis_knit):
438
warnings.warn("KnitVersionedFile.__(): The basis_knit parameter is"
439
" deprecated as of bzr 0.9.",
440
DeprecationWarning, stacklevel=2)
441
if access_mode is None:
443
super(KnitVersionedFile, self).__init__(access_mode)
444
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
445
self.transport = transport
446
self.filename = relpath
447
self.factory = factory or KnitAnnotateFactory()
448
self.writable = (access_mode == 'w')
451
self._max_delta_chain = 200
454
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
455
access_mode, create=create, file_mode=file_mode,
456
create_parent_dir=create_parent_dir, delay_create=delay_create,
460
if access_method is None:
461
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
462
((create and not len(self)) and delay_create), create_parent_dir)
464
_access = access_method
465
if create and not len(self) and not delay_create:
467
self._data = _KnitData(_access)
470
return '%s(%s)' % (self.__class__.__name__,
471
self.transport.abspath(self.filename))
473
def _check_should_delta(self, first_parents):
474
"""Iterate back through the parent listing, looking for a fulltext.
476
This is used when we want to decide whether to add a delta or a new
477
fulltext. It searches for _max_delta_chain parents. When it finds a
478
fulltext parent, it sees if the total size of the deltas leading up to
479
it is large enough to indicate that we want a new full text anyway.
481
Return True if we should create a new delta, False if we should use a
486
delta_parents = first_parents
487
for count in xrange(self._max_delta_chain):
488
parent = delta_parents[0]
489
method = self._index.get_method(parent)
490
index, pos, size = self._index.get_position(parent)
491
if method == 'fulltext':
495
delta_parents = self._index.get_parents(parent)
497
# We couldn't find a fulltext, so we must create a new one
500
return fulltext_size > delta_size
502
def _add_raw_records(self, records, data):
503
"""Add all the records 'records' with data pre-joined in 'data'.
505
:param records: A list of tuples(version_id, options, parents, size).
506
:param data: The data for the records. When it is written, the records
507
are adjusted to have pos pointing into data by the sum of
508
the preceding records sizes.
511
raw_record_sizes = [record[3] for record in records]
512
positions = self._data.add_raw_records(raw_record_sizes, data)
515
for (version_id, options, parents, size), access_memo in zip(
517
index_entries.append((version_id, options, access_memo, parents))
518
if self._data._do_cache:
519
self._data._cache[version_id] = data[offset:offset+size]
521
self._index.add_versions(index_entries)
523
def enable_cache(self):
524
"""Start caching data for this knit"""
525
self._data.enable_cache()
527
def clear_cache(self):
528
"""Clear the data cache only."""
529
self._data.clear_cache()
531
def copy_to(self, name, transport):
532
"""See VersionedFile.copy_to()."""
533
# copy the current index to a temp index to avoid racing with local
535
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
536
self.transport.get(self._index._filename))
538
f = self._data._open_file()
540
transport.put_file(name + DATA_SUFFIX, f)
543
# move the copied index into place
544
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
546
def create_empty(self, name, transport, mode=None):
547
return KnitVersionedFile(name, transport, factory=self.factory,
548
delta=self.delta, create=True)
550
def get_data_stream(self, required_versions):
551
"""Get a data stream for the specified versions.
553
Versions may be returned in any order, not necessarily the order
556
:param required_versions: The exact set of versions to be extracted.
557
Unlike some other knit methods, this is not used to generate a
558
transitive closure, rather it is used precisely as given.
560
:returns: format_signature, list of (version, options, length, parents),
563
required_versions = set([osutils.safe_revision_id(v) for v in
565
# we don't care about inclusions, the caller cares.
566
# but we need to setup a list of records to visit.
567
for version_id in required_versions:
568
if not self.has_version(version_id):
569
raise RevisionNotPresent(version_id, self.filename)
570
# Pick the desired versions out of the index in oldest-to-newest order
572
for version_id in self.versions():
573
if version_id in required_versions:
574
version_list.append(version_id)
576
# create the list of version information for the result
577
copy_queue_records = []
579
result_version_list = []
580
for version_id in version_list:
581
options = self._index.get_options(version_id)
582
parents = self._index.get_parents_with_ghosts(version_id)
583
index_memo = self._index.get_position(version_id)
584
copy_queue_records.append((version_id, index_memo))
585
none, data_pos, data_size = index_memo
586
copy_set.add(version_id)
587
# version, options, length, parents
588
result_version_list.append((version_id, options, data_size,
591
# Read the compressed record data.
593
# From here down to the return should really be logic in the returned
594
# callable -- in a class that adapts read_records_iter_raw to read
597
for (version_id, raw_data), \
598
(version_id2, options, _, parents) in \
599
izip(self._data.read_records_iter_raw(copy_queue_records),
600
result_version_list):
601
assert version_id == version_id2, 'logic error, inconsistent results'
602
raw_datum.append(raw_data)
603
pseudo_file = StringIO(''.join(raw_datum))
606
return pseudo_file.read()
608
return pseudo_file.read(length)
609
return (self.get_format_signature(), result_version_list, read)
611
def _extract_blocks(self, version_id, source, target):
612
if self._index.get_method(version_id) != 'line-delta':
614
parent, sha1, noeol, delta = self.get_delta(version_id)
615
return KnitContent.get_line_delta_blocks(delta, source, target)
617
def get_delta(self, version_id):
618
"""Get a delta for constructing version from some other version."""
619
version_id = osutils.safe_revision_id(version_id)
620
self.check_not_reserved_id(version_id)
621
parents = self.get_parents(version_id)
626
index_memo = self._index.get_position(version_id)
627
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
628
noeol = 'no-eol' in self._index.get_options(version_id)
629
if 'fulltext' == self._index.get_method(version_id):
630
new_content = self.factory.parse_fulltext(data, version_id)
631
if parent is not None:
632
reference_content = self._get_content(parent)
633
old_texts = reference_content.text()
636
new_texts = new_content.text()
637
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
639
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
641
delta = self.factory.parse_line_delta(data, version_id)
642
return parent, sha1, noeol, delta
644
def get_format_signature(self):
645
"""See VersionedFile.get_format_signature()."""
646
if self.factory.annotated:
647
annotated_part = "annotated"
649
annotated_part = "plain"
650
return "knit-%s" % (annotated_part,)
652
def get_graph_with_ghosts(self):
653
"""See VersionedFile.get_graph_with_ghosts()."""
654
graph_items = self._index.get_graph()
655
return dict(graph_items)
657
def get_sha1(self, version_id):
658
return self.get_sha1s([version_id])[0]
660
def get_sha1s(self, version_ids):
661
"""See VersionedFile.get_sha1()."""
662
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
663
record_map = self._get_record_map(version_ids)
664
# record entry 2 is the 'digest'.
665
return [record_map[v][2] for v in version_ids]
669
"""See VersionedFile.get_suffixes()."""
670
return [DATA_SUFFIX, INDEX_SUFFIX]
672
def has_ghost(self, version_id):
673
"""True if there is a ghost reference in the file to version_id."""
674
version_id = osutils.safe_revision_id(version_id)
676
if self.has_version(version_id):
678
# optimisable if needed by memoising the _ghosts set.
679
items = self._index.get_graph()
680
for node, parents in items:
681
for parent in parents:
682
if parent not in self._index._cache:
683
if parent == version_id:
687
def insert_data_stream(self, (format, data_list, reader_callable)):
688
"""Insert knit records from a data stream into this knit.
690
If a version in the stream is already present in this knit, it will not
691
be inserted a second time. It will be checked for consistency with the
692
stored version however, and may cause a KnitCorrupt error to be raised
693
if the data in the stream disagrees with the already stored data.
695
:seealso: get_data_stream
697
if format != self.get_format_signature():
698
trace.mutter('incompatible format signature inserting to %r', self)
699
raise KnitDataStreamIncompatible(
700
format, self.get_format_signature())
702
for version_id, options, length, parents in data_list:
703
if self.has_version(version_id):
704
# First check: the list of parents.
705
my_parents = self.get_parents_with_ghosts(version_id)
706
if my_parents != parents:
707
# XXX: KnitCorrupt is not quite the right exception here.
710
'parents list %r from data stream does not match '
711
'already recorded parents %r for %s'
712
% (parents, my_parents, version_id))
714
# Also check the SHA-1 of the fulltext this content will
716
raw_data = reader_callable(length)
717
my_fulltext_sha1 = self.get_sha1(version_id)
718
df, rec = self._data._parse_record_header(version_id, raw_data)
719
stream_fulltext_sha1 = rec[3]
720
if my_fulltext_sha1 != stream_fulltext_sha1:
721
# Actually, we don't know if it's this knit that's corrupt,
722
# or the data stream we're trying to insert.
724
self.filename, 'sha-1 does not match %s' % version_id)
726
self._add_raw_records(
727
[(version_id, options, parents, length)],
728
reader_callable(length))
731
"""See VersionedFile.versions."""
732
if 'evil' in debug.debug_flags:
733
trace.mutter_callsite(2, "versions scales with size of history")
734
return self._index.get_versions()
736
def has_version(self, version_id):
737
"""See VersionedFile.has_version."""
738
if 'evil' in debug.debug_flags:
739
trace.mutter_callsite(2, "has_version is a LBYL scenario")
740
version_id = osutils.safe_revision_id(version_id)
741
return self._index.has_version(version_id)
743
__contains__ = has_version
745
def _merge_annotations(self, content, parents, parent_texts={},
746
delta=None, annotated=None,
747
left_matching_blocks=None):
748
"""Merge annotations for content. This is done by comparing
749
the annotations based on changed to the text.
751
if left_matching_blocks is not None:
752
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
756
for parent_id in parents:
757
merge_content = self._get_content(parent_id, parent_texts)
758
if (parent_id == parents[0] and delta_seq is not None):
761
seq = patiencediff.PatienceSequenceMatcher(
762
None, merge_content.text(), content.text())
763
for i, j, n in seq.get_matching_blocks():
766
# this appears to copy (origin, text) pairs across to the
767
# new content for any line that matches the last-checked
769
content._lines[j:j+n] = merge_content._lines[i:i+n]
771
if delta_seq is None:
772
reference_content = self._get_content(parents[0], parent_texts)
773
new_texts = content.text()
774
old_texts = reference_content.text()
775
delta_seq = patiencediff.PatienceSequenceMatcher(
776
None, old_texts, new_texts)
777
return self._make_line_delta(delta_seq, content)
779
def _make_line_delta(self, delta_seq, new_content):
780
"""Generate a line delta from delta_seq and new_content."""
782
for op in delta_seq.get_opcodes():
785
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
788
def _get_components_positions(self, version_ids):
789
"""Produce a map of position data for the components of versions.
791
This data is intended to be used for retrieving the knit records.
793
A dict of version_id to (method, data_pos, data_size, next) is
795
method is the way referenced data should be applied.
796
data_pos is the position of the data in the knit.
797
data_size is the size of the data in the knit.
798
next is the build-parent of the version, or None for fulltexts.
801
for version_id in version_ids:
804
while cursor is not None and cursor not in component_data:
805
method = self._index.get_method(cursor)
806
if method == 'fulltext':
809
next = self.get_parents(cursor)[0]
810
index_memo = self._index.get_position(cursor)
811
component_data[cursor] = (method, index_memo, next)
813
return component_data
815
def _get_content(self, version_id, parent_texts={}):
816
"""Returns a content object that makes up the specified
818
cached_version = parent_texts.get(version_id, None)
819
if cached_version is not None:
820
if not self.has_version(version_id):
821
raise RevisionNotPresent(version_id, self.filename)
822
return cached_version
824
text_map, contents_map = self._get_content_maps([version_id])
825
return contents_map[version_id]
827
def _check_versions_present(self, version_ids):
828
"""Check that all specified versions are present."""
829
self._index.check_versions_present(version_ids)
831
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
832
nostore_sha, random_id, check_content):
833
"""See VersionedFile.add_lines_with_ghosts()."""
834
self._check_add(version_id, lines, random_id, check_content)
835
return self._add(version_id, lines, parents, self.delta,
836
parent_texts, None, nostore_sha, random_id)
838
def _add_lines(self, version_id, parents, lines, parent_texts,
839
left_matching_blocks, nostore_sha, random_id, check_content):
840
"""See VersionedFile.add_lines."""
841
self._check_add(version_id, lines, random_id, check_content)
842
self._check_versions_present(parents)
843
return self._add(version_id, lines[:], parents, self.delta,
844
parent_texts, left_matching_blocks, nostore_sha, random_id)
846
def _check_add(self, version_id, lines, random_id, check_content):
847
"""check that version_id and lines are safe to add."""
848
if contains_whitespace(version_id):
849
raise InvalidRevisionId(version_id, self.filename)
850
self.check_not_reserved_id(version_id)
851
# Technically this could be avoided if we are happy to allow duplicate
852
# id insertion when other things than bzr core insert texts, but it
853
# seems useful for folk using the knit api directly to have some safety
854
# blanket that we can disable.
855
if not random_id and self.has_version(version_id):
856
raise RevisionAlreadyPresent(version_id, self.filename)
858
self._check_lines_not_unicode(lines)
859
self._check_lines_are_lines(lines)
861
def _add(self, version_id, lines, parents, delta, parent_texts,
862
left_matching_blocks, nostore_sha, random_id):
863
"""Add a set of lines on top of version specified by parents.
865
If delta is true, compress the text as a line-delta against
868
Any versions not present will be converted into ghosts.
870
# first thing, if the content is something we don't need to store, find
872
line_bytes = ''.join(lines)
873
digest = sha_string(line_bytes)
874
if nostore_sha == digest:
875
raise errors.ExistingContent
878
if parent_texts is None:
880
for parent in parents:
881
if self.has_version(parent):
882
present_parents.append(parent)
884
# can only compress against the left most present parent.
886
(len(present_parents) == 0 or
887
present_parents[0] != parents[0])):
890
text_length = len(line_bytes)
893
if lines[-1][-1] != '\n':
894
# copy the contents of lines.
896
options.append('no-eol')
897
lines[-1] = lines[-1] + '\n'
900
# To speed the extract of texts the delta chain is limited
901
# to a fixed number of deltas. This should minimize both
902
# I/O and the time spend applying deltas.
903
delta = self._check_should_delta(present_parents)
905
assert isinstance(version_id, str)
906
content = self.factory.make(lines, version_id)
907
if delta or (self.factory.annotated and len(present_parents) > 0):
908
# Merge annotations from parent texts if needed.
909
delta_hunks = self._merge_annotations(content, present_parents,
910
parent_texts, delta, self.factory.annotated,
911
left_matching_blocks)
914
options.append('line-delta')
915
store_lines = self.factory.lower_line_delta(delta_hunks)
916
size, bytes = self._data._record_to_data(version_id, digest,
919
options.append('fulltext')
920
# get mixed annotation + content and feed it into the
922
store_lines = self.factory.lower_fulltext(content)
923
size, bytes = self._data._record_to_data(version_id, digest,
926
access_memo = self._data.add_raw_records([size], bytes)[0]
927
self._index.add_versions(
928
((version_id, options, access_memo, parents),),
930
return digest, text_length, content
932
def check(self, progress_bar=None):
933
"""See VersionedFile.check()."""
935
def _clone_text(self, new_version_id, old_version_id, parents):
936
"""See VersionedFile.clone_text()."""
937
# FIXME RBC 20060228 make fast by only inserting an index with null
939
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
941
def get_lines(self, version_id):
942
"""See VersionedFile.get_lines()."""
943
return self.get_line_list([version_id])[0]
945
def _get_record_map(self, version_ids):
946
"""Produce a dictionary of knit records.
948
The keys are version_ids, the values are tuples of (method, content,
950
method is the way the content should be applied.
951
content is a KnitContent object.
952
digest is the SHA1 digest of this version id after all steps are done
953
next is the build-parent of the version, i.e. the leftmost ancestor.
954
If the method is fulltext, next will be None.
956
position_map = self._get_components_positions(version_ids)
957
# c = component_id, m = method, i_m = index_memo, n = next
958
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
960
for component_id, content, digest in \
961
self._data.read_records_iter(records):
962
method, index_memo, next = position_map[component_id]
963
record_map[component_id] = method, content, digest, next
967
def get_text(self, version_id):
968
"""See VersionedFile.get_text"""
969
return self.get_texts([version_id])[0]
971
def get_texts(self, version_ids):
972
return [''.join(l) for l in self.get_line_list(version_ids)]
974
def get_line_list(self, version_ids):
975
"""Return the texts of listed versions as a list of strings."""
976
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
977
for version_id in version_ids:
978
self.check_not_reserved_id(version_id)
979
text_map, content_map = self._get_content_maps(version_ids)
980
return [text_map[v] for v in version_ids]
982
_get_lf_split_line_list = get_line_list
984
def _get_content_maps(self, version_ids):
985
"""Produce maps of text and KnitContents
987
:return: (text_map, content_map) where text_map contains the texts for
988
the requested versions and content_map contains the KnitContents.
989
Both dicts take version_ids as their keys.
991
for version_id in version_ids:
992
if not self.has_version(version_id):
993
raise RevisionNotPresent(version_id, self.filename)
994
record_map = self._get_record_map(version_ids)
999
for version_id in version_ids:
1002
while cursor is not None:
1003
method, data, digest, next = record_map[cursor]
1004
components.append((cursor, method, data, digest))
1005
if cursor in content_map:
1010
for component_id, method, data, digest in reversed(components):
1011
if component_id in content_map:
1012
content = content_map[component_id]
1014
if method == 'fulltext':
1015
assert content is None
1016
content = self.factory.parse_fulltext(data, version_id)
1017
elif method == 'line-delta':
1018
delta = self.factory.parse_line_delta(data, version_id)
1019
content = content.copy()
1020
content._lines = self._apply_delta(content._lines,
1022
content_map[component_id] = content
1024
if 'no-eol' in self._index.get_options(version_id):
1025
content = content.copy()
1026
content.strip_last_line_newline()
1027
final_content[version_id] = content
1029
# digest here is the digest from the last applied component.
1030
text = content.text()
1031
if sha_strings(text) != digest:
1032
raise KnitCorrupt(self.filename,
1033
'sha-1 does not match %s' % version_id)
1035
text_map[version_id] = text
1036
return text_map, final_content
1039
def _apply_delta(lines, delta):
1040
"""Apply delta to lines."""
1043
for start, end, count, delta_lines in delta:
1044
lines[offset+start:offset+end] = delta_lines
1045
offset = offset + (start - end) + count
1048
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1050
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1051
if version_ids is None:
1052
version_ids = self.versions()
1054
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
1056
pb = progress.DummyProgress()
1057
# we don't care about inclusions, the caller cares.
1058
# but we need to setup a list of records to visit.
1059
# we need version_id, position, length
1060
version_id_records = []
1061
requested_versions = set(version_ids)
1062
# filter for available versions
1063
for version_id in requested_versions:
1064
if not self.has_version(version_id):
1065
raise RevisionNotPresent(version_id, self.filename)
1066
# get a in-component-order queue:
1067
for version_id in self.versions():
1068
if version_id in requested_versions:
1069
index_memo = self._index.get_position(version_id)
1070
version_id_records.append((version_id, index_memo))
1072
total = len(version_id_records)
1073
for version_idx, (version_id, data, sha_value) in \
1074
enumerate(self._data.read_records_iter(version_id_records)):
1075
pb.update('Walking content.', version_idx, total)
1076
method = self._index.get_method(version_id)
1078
assert method in ('fulltext', 'line-delta')
1079
if method == 'fulltext':
1080
line_iterator = self.factory.get_fulltext_content(data)
1082
line_iterator = self.factory.get_linedelta_content(data)
1083
for line in line_iterator:
1086
pb.update('Walking content.', total, total)
1088
def iter_parents(self, version_ids):
1089
"""Iterate through the parents for many version ids.
1091
:param version_ids: An iterable yielding version_ids.
1092
:return: An iterator that yields (version_id, parents). Requested
1093
version_ids not present in the versioned file are simply skipped.
1094
The order is undefined, allowing for different optimisations in
1095
the underlying implementation.
1097
version_ids = [osutils.safe_revision_id(version_id) for
1098
version_id in version_ids]
1099
return self._index.iter_parents(version_ids)
1101
def num_versions(self):
1102
"""See VersionedFile.num_versions()."""
1103
return self._index.num_versions()
1105
__len__ = num_versions
1107
def annotate_iter(self, version_id):
1108
"""See VersionedFile.annotate_iter."""
1109
version_id = osutils.safe_revision_id(version_id)
1110
return self.factory.annotate_iter(self, version_id)
1112
def get_parents(self, version_id):
1113
"""See VersionedFile.get_parents."""
1116
# 52554 calls in 1264 872 internal down from 3674
1117
version_id = osutils.safe_revision_id(version_id)
1119
return self._index.get_parents(version_id)
1121
raise RevisionNotPresent(version_id, self.filename)
1123
def get_parents_with_ghosts(self, version_id):
1124
"""See VersionedFile.get_parents."""
1125
version_id = osutils.safe_revision_id(version_id)
1127
return self._index.get_parents_with_ghosts(version_id)
1129
raise RevisionNotPresent(version_id, self.filename)
1131
def get_ancestry(self, versions, topo_sorted=True):
1132
"""See VersionedFile.get_ancestry."""
1133
if isinstance(versions, basestring):
1134
versions = [versions]
1137
versions = [osutils.safe_revision_id(v) for v in versions]
1138
return self._index.get_ancestry(versions, topo_sorted)
1140
def get_ancestry_with_ghosts(self, versions):
1141
"""See VersionedFile.get_ancestry_with_ghosts."""
1142
if isinstance(versions, basestring):
1143
versions = [versions]
1146
versions = [osutils.safe_revision_id(v) for v in versions]
1147
return self._index.get_ancestry_with_ghosts(versions)
1149
def plan_merge(self, ver_a, ver_b):
1150
"""See VersionedFile.plan_merge."""
1151
ver_a = osutils.safe_revision_id(ver_a)
1152
ver_b = osutils.safe_revision_id(ver_b)
1153
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1155
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1156
annotated_a = self.annotate(ver_a)
1157
annotated_b = self.annotate(ver_b)
1158
return merge._plan_annotate_merge(annotated_a, annotated_b,
1159
ancestors_a, ancestors_b)
1162
class _KnitComponentFile(object):
1163
"""One of the files used to implement a knit database"""
1165
def __init__(self, transport, filename, mode, file_mode=None,
1166
create_parent_dir=False, dir_mode=None):
1167
self._transport = transport
1168
self._filename = filename
1170
self._file_mode = file_mode
1171
self._dir_mode = dir_mode
1172
self._create_parent_dir = create_parent_dir
1173
self._need_to_create = False
1175
def _full_path(self):
1176
"""Return the full path to this file."""
1177
return self._transport.base + self._filename
1179
def check_header(self, fp):
1180
line = fp.readline()
1182
# An empty file can actually be treated as though the file doesn't
1184
raise errors.NoSuchFile(self._full_path())
1185
if line != self.HEADER:
1186
raise KnitHeaderError(badline=line,
1187
filename=self._transport.abspath(self._filename))
1190
return '%s(%s)' % (self.__class__.__name__, self._filename)
1193
class _KnitIndex(_KnitComponentFile):
1194
"""Manages knit index file.
1196
The index is already kept in memory and read on startup, to enable
1197
fast lookups of revision information. The cursor of the index
1198
file is always pointing to the end, making it easy to append
1201
_cache is a cache for fast mapping from version id to a Index
1204
_history is a cache for fast mapping from indexes to version ids.
1206
The index data format is dictionary compressed when it comes to
1207
parent references; a index entry may only have parents that with a
1208
lover index number. As a result, the index is topological sorted.
1210
Duplicate entries may be written to the index for a single version id
1211
if this is done then the latter one completely replaces the former:
1212
this allows updates to correct version and parent information.
1213
Note that the two entries may share the delta, and that successive
1214
annotations and references MUST point to the first entry.
1216
The index file on disc contains a header, followed by one line per knit
1217
record. The same revision can be present in an index file more than once.
1218
The first occurrence gets assigned a sequence number starting from 0.
1220
The format of a single line is
1221
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1222
REVISION_ID is a utf8-encoded revision id
1223
FLAGS is a comma separated list of flags about the record. Values include
1224
no-eol, line-delta, fulltext.
1225
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1226
that the the compressed data starts at.
1227
LENGTH is the ascii representation of the length of the data file.
1228
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1230
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1231
revision id already in the knit that is a parent of REVISION_ID.
1232
The ' :' marker is the end of record marker.
1235
when a write is interrupted to the index file, it will result in a line
1236
that does not end in ' :'. If the ' :' is not present at the end of a line,
1237
or at the end of the file, then the record that is missing it will be
1238
ignored by the parser.
1240
When writing new records to the index file, the data is preceded by '\n'
1241
to ensure that records always start on new lines even if the last write was
1242
interrupted. As a result its normal for the last line in the index to be
1243
missing a trailing newline. One can be added with no harmful effects.
1246
HEADER = "# bzr knit index 8\n"
1248
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1249
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1251
def _cache_version(self, version_id, options, pos, size, parents):
1252
"""Cache a version record in the history array and index cache.
1254
This is inlined into _load_data for performance. KEEP IN SYNC.
1255
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1258
# only want the _history index to reference the 1st index entry
1260
if version_id not in self._cache:
1261
index = len(self._history)
1262
self._history.append(version_id)
1264
index = self._cache[version_id][5]
1265
self._cache[version_id] = (version_id,
1272
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1273
create_parent_dir=False, delay_create=False, dir_mode=None):
1274
_KnitComponentFile.__init__(self, transport, filename, mode,
1275
file_mode=file_mode,
1276
create_parent_dir=create_parent_dir,
1279
# position in _history is the 'official' index for a revision
1280
# but the values may have come from a newer entry.
1281
# so - wc -l of a knit index is != the number of unique names
1285
fp = self._transport.get(self._filename)
1287
# _load_data may raise NoSuchFile if the target knit is
1289
_load_data(self, fp)
1293
if mode != 'w' or not create:
1296
self._need_to_create = True
1298
self._transport.put_bytes_non_atomic(
1299
self._filename, self.HEADER, mode=self._file_mode)
1301
def get_graph(self):
1302
"""Return a list of the node:parents lists from this knit index."""
1303
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1305
def get_ancestry(self, versions, topo_sorted=True):
1306
"""See VersionedFile.get_ancestry."""
1307
# get a graph of all the mentioned versions:
1309
pending = set(versions)
1312
version = pending.pop()
1315
parents = [p for p in cache[version][4] if p in cache]
1317
raise RevisionNotPresent(version, self._filename)
1318
# if not completed and not a ghost
1319
pending.update([p for p in parents if p not in graph])
1320
graph[version] = parents
1323
return topo_sort(graph.items())
1325
def get_ancestry_with_ghosts(self, versions):
1326
"""See VersionedFile.get_ancestry_with_ghosts."""
1327
# get a graph of all the mentioned versions:
1328
self.check_versions_present(versions)
1331
pending = set(versions)
1333
version = pending.pop()
1335
parents = cache[version][4]
1341
pending.update([p for p in parents if p not in graph])
1342
graph[version] = parents
1343
return topo_sort(graph.items())
1345
def iter_parents(self, version_ids):
1346
"""Iterate through the parents for many version ids.
1348
:param version_ids: An iterable yielding version_ids.
1349
:return: An iterator that yields (version_id, parents). Requested
1350
version_ids not present in the versioned file are simply skipped.
1351
The order is undefined, allowing for different optimisations in
1352
the underlying implementation.
1354
for version_id in version_ids:
1356
yield version_id, tuple(self.get_parents(version_id))
1360
def num_versions(self):
1361
return len(self._history)
1363
__len__ = num_versions
1365
def get_versions(self):
1366
"""Get all the versions in the file. not topologically sorted."""
1367
return self._history
1369
def _version_list_to_index(self, versions):
1372
for version in versions:
1373
if version in cache:
1374
# -- inlined lookup() --
1375
result_list.append(str(cache[version][5]))
1376
# -- end lookup () --
1378
result_list.append('.' + version)
1379
return ' '.join(result_list)
1381
def add_version(self, version_id, options, index_memo, parents):
1382
"""Add a version record to the index."""
1383
self.add_versions(((version_id, options, index_memo, parents),))
1385
def add_versions(self, versions, random_id=False):
1386
"""Add multiple versions to the index.
1388
:param versions: a list of tuples:
1389
(version_id, options, pos, size, parents).
1390
:param random_id: If True the ids being added were randomly generated
1391
and no check for existence will be performed.
1394
orig_history = self._history[:]
1395
orig_cache = self._cache.copy()
1398
for version_id, options, (index, pos, size), parents in versions:
1399
line = "\n%s %s %s %s %s :" % (version_id,
1403
self._version_list_to_index(parents))
1404
assert isinstance(line, str), \
1405
'content must be utf-8 encoded: %r' % (line,)
1407
self._cache_version(version_id, options, pos, size, parents)
1408
if not self._need_to_create:
1409
self._transport.append_bytes(self._filename, ''.join(lines))
1412
sio.write(self.HEADER)
1413
sio.writelines(lines)
1415
self._transport.put_file_non_atomic(self._filename, sio,
1416
create_parent_dir=self._create_parent_dir,
1417
mode=self._file_mode,
1418
dir_mode=self._dir_mode)
1419
self._need_to_create = False
1421
# If any problems happen, restore the original values and re-raise
1422
self._history = orig_history
1423
self._cache = orig_cache
1426
def has_version(self, version_id):
1427
"""True if the version is in the index."""
1428
return version_id in self._cache
1430
def get_position(self, version_id):
1431
"""Return details needed to access the version.
1433
.kndx indices do not support split-out data, so return None for the
1436
:return: a tuple (None, data position, size) to hand to the access
1437
logic to get the record.
1439
entry = self._cache[version_id]
1440
return None, entry[2], entry[3]
1442
def get_method(self, version_id):
1443
"""Return compression method of specified version."""
1444
options = self._cache[version_id][1]
1445
if 'fulltext' in options:
1448
if 'line-delta' not in options:
1449
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1452
def get_options(self, version_id):
1453
"""Return a string represention options.
1457
return self._cache[version_id][1]
1459
def get_parents(self, version_id):
1460
"""Return parents of specified version ignoring ghosts."""
1461
return [parent for parent in self._cache[version_id][4]
1462
if parent in self._cache]
1464
def get_parents_with_ghosts(self, version_id):
1465
"""Return parents of specified version with ghosts."""
1466
return self._cache[version_id][4]
1468
def check_versions_present(self, version_ids):
1469
"""Check that all specified versions are present."""
1471
for version_id in version_ids:
1472
if version_id not in cache:
1473
raise RevisionNotPresent(version_id, self._filename)
1476
class KnitGraphIndex(object):
1477
"""A knit index that builds on GraphIndex."""
1479
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1480
"""Construct a KnitGraphIndex on a graph_index.
1482
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1483
:param deltas: Allow delta-compressed records.
1484
:param add_callback: If not None, allow additions to the index and call
1485
this callback with a list of added GraphIndex nodes:
1486
[(node, value, node_refs), ...]
1487
:param parents: If True, record knits parents, if not do not record
1490
self._graph_index = graph_index
1491
self._deltas = deltas
1492
self._add_callback = add_callback
1493
self._parents = parents
1494
if deltas and not parents:
1495
raise KnitCorrupt(self, "Cannot do delta compression without "
1498
def _get_entries(self, keys, check_present=False):
1499
"""Get the entries for keys.
1501
:param keys: An iterable of index keys, - 1-tuples.
1506
for node in self._graph_index.iter_entries(keys):
1508
found_keys.add(node[1])
1510
# adapt parentless index to the rest of the code.
1511
for node in self._graph_index.iter_entries(keys):
1512
yield node[0], node[1], node[2], ()
1513
found_keys.add(node[1])
1515
missing_keys = keys.difference(found_keys)
1517
raise RevisionNotPresent(missing_keys.pop(), self)
1519
def _present_keys(self, version_ids):
1521
node[1] for node in self._get_entries(version_ids)])
1523
def _parentless_ancestry(self, versions):
1524
"""Honour the get_ancestry API for parentless knit indices."""
1525
wanted_keys = self._version_ids_to_keys(versions)
1526
present_keys = self._present_keys(wanted_keys)
1527
missing = set(wanted_keys).difference(present_keys)
1529
raise RevisionNotPresent(missing.pop(), self)
1530
return list(self._keys_to_version_ids(present_keys))
1532
def get_ancestry(self, versions, topo_sorted=True):
1533
"""See VersionedFile.get_ancestry."""
1534
if not self._parents:
1535
return self._parentless_ancestry(versions)
1536
# XXX: This will do len(history) index calls - perhaps
1537
# it should be altered to be a index core feature?
1538
# get a graph of all the mentioned versions:
1541
versions = self._version_ids_to_keys(versions)
1542
pending = set(versions)
1544
# get all pending nodes
1545
this_iteration = pending
1546
new_nodes = self._get_entries(this_iteration)
1549
for (index, key, value, node_refs) in new_nodes:
1550
# dont ask for ghosties - otherwise
1551
# we we can end up looping with pending
1552
# being entirely ghosted.
1553
graph[key] = [parent for parent in node_refs[0]
1554
if parent not in ghosts]
1556
for parent in graph[key]:
1557
# dont examine known nodes again
1562
ghosts.update(this_iteration.difference(found))
1563
if versions.difference(graph):
1564
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1566
result_keys = topo_sort(graph.items())
1568
result_keys = graph.iterkeys()
1569
return [key[0] for key in result_keys]
1571
def get_ancestry_with_ghosts(self, versions):
1572
"""See VersionedFile.get_ancestry."""
1573
if not self._parents:
1574
return self._parentless_ancestry(versions)
1575
# XXX: This will do len(history) index calls - perhaps
1576
# it should be altered to be a index core feature?
1577
# get a graph of all the mentioned versions:
1579
versions = self._version_ids_to_keys(versions)
1580
pending = set(versions)
1582
# get all pending nodes
1583
this_iteration = pending
1584
new_nodes = self._get_entries(this_iteration)
1586
for (index, key, value, node_refs) in new_nodes:
1587
graph[key] = node_refs[0]
1589
for parent in graph[key]:
1590
# dont examine known nodes again
1594
missing_versions = this_iteration.difference(graph)
1595
missing_needed = versions.intersection(missing_versions)
1597
raise RevisionNotPresent(missing_needed.pop(), self)
1598
for missing_version in missing_versions:
1599
# add a key, no parents
1600
graph[missing_version] = []
1601
pending.discard(missing_version) # don't look for it
1602
result_keys = topo_sort(graph.items())
1603
return [key[0] for key in result_keys]
1605
def get_graph(self):
1606
"""Return a list of the node:parents lists from this knit index."""
1607
if not self._parents:
1608
return [(key, ()) for key in self.get_versions()]
1610
for index, key, value, refs in self._graph_index.iter_all_entries():
1611
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1614
def iter_parents(self, version_ids):
1615
"""Iterate through the parents for many version ids.
1617
:param version_ids: An iterable yielding version_ids.
1618
:return: An iterator that yields (version_id, parents). Requested
1619
version_ids not present in the versioned file are simply skipped.
1620
The order is undefined, allowing for different optimisations in
1621
the underlying implementation.
1624
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1626
present_parents = set()
1627
for node in all_nodes:
1628
all_parents.update(node[3][0])
1629
# any node we are querying must be present
1630
present_parents.add(node[1])
1631
unknown_parents = all_parents.difference(present_parents)
1632
present_parents.update(self._present_keys(unknown_parents))
1633
for node in all_nodes:
1635
for parent in node[3][0]:
1636
if parent in present_parents:
1637
parents.append(parent[0])
1638
yield node[1][0], tuple(parents)
1640
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1641
yield node[1][0], ()
1643
def num_versions(self):
1644
return len(list(self._graph_index.iter_all_entries()))
1646
__len__ = num_versions
1648
def get_versions(self):
1649
"""Get all the versions in the file. not topologically sorted."""
1650
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1652
def has_version(self, version_id):
1653
"""True if the version is in the index."""
1654
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1656
def _keys_to_version_ids(self, keys):
1657
return tuple(key[0] for key in keys)
1659
def get_position(self, version_id):
1660
"""Return details needed to access the version.
1662
:return: a tuple (index, data position, size) to hand to the access
1663
logic to get the record.
1665
node = self._get_node(version_id)
1666
bits = node[2][1:].split(' ')
1667
return node[0], int(bits[0]), int(bits[1])
1669
def get_method(self, version_id):
1670
"""Return compression method of specified version."""
1671
if not self._deltas:
1673
return self._parent_compression(self._get_node(version_id)[3][1])
1675
def _parent_compression(self, reference_list):
1676
# use the second reference list to decide if this is delta'd or not.
1677
if len(reference_list):
1682
def _get_node(self, version_id):
1683
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1685
def get_options(self, version_id):
1686
"""Return a string represention options.
1690
node = self._get_node(version_id)
1691
if not self._deltas:
1692
options = ['fulltext']
1694
options = [self._parent_compression(node[3][1])]
1695
if node[2][0] == 'N':
1696
options.append('no-eol')
1699
def get_parents(self, version_id):
1700
"""Return parents of specified version ignoring ghosts."""
1701
parents = list(self.iter_parents([version_id]))
1704
raise errors.RevisionNotPresent(version_id, self)
1705
return parents[0][1]
1707
def get_parents_with_ghosts(self, version_id):
1708
"""Return parents of specified version with ghosts."""
1709
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1710
check_present=True))
1711
if not self._parents:
1713
return self._keys_to_version_ids(nodes[0][3][0])
1715
def check_versions_present(self, version_ids):
1716
"""Check that all specified versions are present."""
1717
keys = self._version_ids_to_keys(version_ids)
1718
present = self._present_keys(keys)
1719
missing = keys.difference(present)
1721
raise RevisionNotPresent(missing.pop(), self)
1723
def add_version(self, version_id, options, access_memo, parents):
1724
"""Add a version record to the index."""
1725
return self.add_versions(((version_id, options, access_memo, parents),))
1727
def add_versions(self, versions, random_id=False):
1728
"""Add multiple versions to the index.
1730
This function does not insert data into the Immutable GraphIndex
1731
backing the KnitGraphIndex, instead it prepares data for insertion by
1732
the caller and checks that it is safe to insert then calls
1733
self._add_callback with the prepared GraphIndex nodes.
1735
:param versions: a list of tuples:
1736
(version_id, options, pos, size, parents).
1737
:param random_id: If True the ids being added were randomly generated
1738
and no check for existence will be performed.
1740
if not self._add_callback:
1741
raise errors.ReadOnlyError(self)
1742
# we hope there are no repositories with inconsistent parentage
1747
for (version_id, options, access_memo, parents) in versions:
1748
index, pos, size = access_memo
1749
key = (version_id, )
1750
parents = tuple((parent, ) for parent in parents)
1751
if 'no-eol' in options:
1755
value += "%d %d" % (pos, size)
1756
if not self._deltas:
1757
if 'line-delta' in options:
1758
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1761
if 'line-delta' in options:
1762
node_refs = (parents, (parents[0],))
1764
node_refs = (parents, ())
1766
node_refs = (parents, )
1769
raise KnitCorrupt(self, "attempt to add node with parents "
1770
"in parentless index.")
1772
keys[key] = (value, node_refs)
1774
present_nodes = self._get_entries(keys)
1775
for (index, key, value, node_refs) in present_nodes:
1776
if (value, node_refs) != keys[key]:
1777
raise KnitCorrupt(self, "inconsistent details in add_versions"
1778
": %s %s" % ((value, node_refs), keys[key]))
1782
for key, (value, node_refs) in keys.iteritems():
1783
result.append((key, value, node_refs))
1785
for key, (value, node_refs) in keys.iteritems():
1786
result.append((key, value))
1787
self._add_callback(result)
1789
def _version_ids_to_keys(self, version_ids):
1790
return set((version_id, ) for version_id in version_ids)
1793
class _KnitAccess(object):
1794
"""Access to knit records in a .knit file."""
1796
def __init__(self, transport, filename, _file_mode, _dir_mode,
1797
_need_to_create, _create_parent_dir):
1798
"""Create a _KnitAccess for accessing and inserting data.
1800
:param transport: The transport the .knit is located on.
1801
:param filename: The filename of the .knit.
1803
self._transport = transport
1804
self._filename = filename
1805
self._file_mode = _file_mode
1806
self._dir_mode = _dir_mode
1807
self._need_to_create = _need_to_create
1808
self._create_parent_dir = _create_parent_dir
1810
def add_raw_records(self, sizes, raw_data):
1811
"""Add raw knit bytes to a storage area.
1813
The data is spooled to whereever the access method is storing data.
1815
:param sizes: An iterable containing the size of each raw data segment.
1816
:param raw_data: A bytestring containing the data.
1817
:return: A list of memos to retrieve the record later. Each memo is a
1818
tuple - (index, pos, length), where the index field is always None
1819
for the .knit access method.
1821
assert type(raw_data) == str, \
1822
'data must be plain bytes was %s' % type(raw_data)
1823
if not self._need_to_create:
1824
base = self._transport.append_bytes(self._filename, raw_data)
1826
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1827
create_parent_dir=self._create_parent_dir,
1828
mode=self._file_mode,
1829
dir_mode=self._dir_mode)
1830
self._need_to_create = False
1834
result.append((None, base, size))
1839
"""IFF this data access has its own storage area, initialise it.
1843
self._transport.put_bytes_non_atomic(self._filename, '',
1844
mode=self._file_mode)
1846
def open_file(self):
1847
"""IFF this data access can be represented as a single file, open it.
1849
For knits that are not mapped to a single file on disk this will
1852
:return: None or a file handle.
1855
return self._transport.get(self._filename)
1860
def get_raw_records(self, memos_for_retrieval):
1861
"""Get the raw bytes for a records.
1863
:param memos_for_retrieval: An iterable containing the (index, pos,
1864
length) memo for retrieving the bytes. The .knit method ignores
1865
the index as there is always only a single file.
1866
:return: An iterator over the bytes of the records.
1868
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1869
for pos, data in self._transport.readv(self._filename, read_vector):
1873
class _PackAccess(object):
1874
"""Access to knit records via a collection of packs."""
1876
def __init__(self, index_to_packs, writer=None):
1877
"""Create a _PackAccess object.
1879
:param index_to_packs: A dict mapping index objects to the transport
1880
and file names for obtaining data.
1881
:param writer: A tuple (pack.ContainerWriter, write_index) which
1882
contains the pack to write, and the index that reads from it will
1886
self.container_writer = writer[0]
1887
self.write_index = writer[1]
1889
self.container_writer = None
1890
self.write_index = None
1891
self.indices = index_to_packs
1893
def add_raw_records(self, sizes, raw_data):
1894
"""Add raw knit bytes to a storage area.
1896
The data is spooled to the container writer in one bytes-record per
1899
:param sizes: An iterable containing the size of each raw data segment.
1900
:param raw_data: A bytestring containing the data.
1901
:return: A list of memos to retrieve the record later. Each memo is a
1902
tuple - (index, pos, length), where the index field is the
1903
write_index object supplied to the PackAccess object.
1905
assert type(raw_data) == str, \
1906
'data must be plain bytes was %s' % type(raw_data)
1910
p_offset, p_length = self.container_writer.add_bytes_record(
1911
raw_data[offset:offset+size], [])
1913
result.append((self.write_index, p_offset, p_length))
1917
"""Pack based knits do not get individually created."""
1919
def get_raw_records(self, memos_for_retrieval):
1920
"""Get the raw bytes for a records.
1922
:param memos_for_retrieval: An iterable containing the (index, pos,
1923
length) memo for retrieving the bytes. The Pack access method
1924
looks up the pack to use for a given record in its index_to_pack
1926
:return: An iterator over the bytes of the records.
1928
# first pass, group into same-index requests
1930
current_index = None
1931
for (index, offset, length) in memos_for_retrieval:
1932
if current_index == index:
1933
current_list.append((offset, length))
1935
if current_index is not None:
1936
request_lists.append((current_index, current_list))
1937
current_index = index
1938
current_list = [(offset, length)]
1939
# handle the last entry
1940
if current_index is not None:
1941
request_lists.append((current_index, current_list))
1942
for index, offsets in request_lists:
1943
transport, path = self.indices[index]
1944
reader = pack.make_readv_reader(transport, path, offsets)
1945
for names, read_func in reader.iter_records():
1946
yield read_func(None)
1948
def open_file(self):
1949
"""Pack based knits have no single file."""
1952
def set_writer(self, writer, index, (transport, packname)):
1953
"""Set a writer to use for adding data."""
1954
self.indices[index] = (transport, packname)
1955
self.container_writer = writer
1956
self.write_index = index
1959
class _KnitData(object):
1960
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1962
The KnitData class provides the logic for parsing and using knit records,
1963
making use of an access method for the low level read and write operations.
1966
def __init__(self, access):
1967
"""Create a KnitData object.
1969
:param access: The access method to use. Access methods such as
1970
_KnitAccess manage the insertion of raw records and the subsequent
1971
retrieval of the same.
1973
self._access = access
1974
self._checked = False
1975
# TODO: jam 20060713 conceptually, this could spill to disk
1976
# if the cached size gets larger than a certain amount
1977
# but it complicates the model a bit, so for now just use
1978
# a simple dictionary
1980
self._do_cache = False
1982
def enable_cache(self):
1983
"""Enable caching of reads."""
1984
self._do_cache = True
1986
def clear_cache(self):
1987
"""Clear the record cache."""
1988
self._do_cache = False
1991
def _open_file(self):
1992
return self._access.open_file()
1994
def _record_to_data(self, version_id, digest, lines):
1995
"""Convert version_id, digest, lines into a raw data block.
1997
:return: (len, a StringIO instance with the raw data ready to read.)
1999
bytes = (''.join(chain(
2000
["version %s %d %s\n" % (version_id,
2004
["end %s\n" % version_id])))
2005
assert bytes.__class__ == str
2006
compressed_bytes = bytes_to_gzip(bytes)
2007
return len(compressed_bytes), compressed_bytes
2009
def add_raw_records(self, sizes, raw_data):
2010
"""Append a prepared record to the data file.
2012
:param sizes: An iterable containing the size of each raw data segment.
2013
:param raw_data: A bytestring containing the data.
2014
:return: a list of index data for the way the data was stored.
2015
See the access method add_raw_records documentation for more
2018
return self._access.add_raw_records(sizes, raw_data)
2020
def _parse_record_header(self, version_id, raw_data):
2021
"""Parse a record header for consistency.
2023
:return: the header and the decompressor stream.
2024
as (stream, header_record)
2026
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2028
rec = self._check_header(version_id, df.readline())
2029
except Exception, e:
2030
raise KnitCorrupt(self._access,
2031
"While reading {%s} got %s(%s)"
2032
% (version_id, e.__class__.__name__, str(e)))
2035
def _check_header(self, version_id, line):
2038
raise KnitCorrupt(self._access,
2039
'unexpected number of elements in record header')
2040
if rec[1] != version_id:
2041
raise KnitCorrupt(self._access,
2042
'unexpected version, wanted %r, got %r'
2043
% (version_id, rec[1]))
2046
def _parse_record(self, version_id, data):
2048
# 4168 calls in 2880 217 internal
2049
# 4168 calls to _parse_record_header in 2121
2050
# 4168 calls to readlines in 330
2051
df = GzipFile(mode='rb', fileobj=StringIO(data))
2054
record_contents = df.readlines()
2055
except Exception, e:
2056
raise KnitCorrupt(self._access,
2057
"While reading {%s} got %s(%s)"
2058
% (version_id, e.__class__.__name__, str(e)))
2059
header = record_contents.pop(0)
2060
rec = self._check_header(version_id, header)
2062
last_line = record_contents.pop()
2063
if len(record_contents) != int(rec[2]):
2064
raise KnitCorrupt(self._access,
2065
'incorrect number of lines %s != %s'
2067
% (len(record_contents), int(rec[2]),
2069
if last_line != 'end %s\n' % rec[1]:
2070
raise KnitCorrupt(self._access,
2071
'unexpected version end line %r, wanted %r'
2072
% (last_line, version_id))
2074
return record_contents, rec[3]
2076
def read_records_iter_raw(self, records):
2077
"""Read text records from data file and yield raw data.
2079
This unpacks enough of the text record to validate the id is
2080
as expected but thats all.
2082
# setup an iterator of the external records:
2083
# uses readv so nice and fast we hope.
2085
# grab the disk data needed.
2087
# Don't check _cache if it is empty
2088
needed_offsets = [index_memo for version_id, index_memo
2090
if version_id not in self._cache]
2092
needed_offsets = [index_memo for version_id, index_memo
2095
raw_records = self._access.get_raw_records(needed_offsets)
2097
for version_id, index_memo in records:
2098
if version_id in self._cache:
2099
# This data has already been validated
2100
data = self._cache[version_id]
2102
data = raw_records.next()
2104
self._cache[version_id] = data
2106
# validate the header
2107
df, rec = self._parse_record_header(version_id, data)
2109
yield version_id, data
2111
def read_records_iter(self, records):
2112
"""Read text records from data file and yield result.
2114
The result will be returned in whatever is the fastest to read.
2115
Not by the order requested. Also, multiple requests for the same
2116
record will only yield 1 response.
2117
:param records: A list of (version_id, pos, len) entries
2118
:return: Yields (version_id, contents, digest) in the order
2119
read, not the order requested
2125
# Skip records we have alread seen
2126
yielded_records = set()
2127
needed_records = set()
2128
for record in records:
2129
if record[0] in self._cache:
2130
if record[0] in yielded_records:
2132
yielded_records.add(record[0])
2133
data = self._cache[record[0]]
2134
content, digest = self._parse_record(record[0], data)
2135
yield (record[0], content, digest)
2137
needed_records.add(record)
2138
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2140
needed_records = sorted(set(records), key=operator.itemgetter(1))
2142
if not needed_records:
2145
# The transport optimizes the fetching as well
2146
# (ie, reads continuous ranges.)
2147
raw_data = self._access.get_raw_records(
2148
[index_memo for version_id, index_memo in needed_records])
2150
for (version_id, index_memo), data in \
2151
izip(iter(needed_records), raw_data):
2152
content, digest = self._parse_record(version_id, data)
2154
self._cache[version_id] = data
2155
yield version_id, content, digest
2157
def read_records(self, records):
2158
"""Read records into a dictionary."""
2160
for record_id, content, digest in \
2161
self.read_records_iter(records):
2162
components[record_id] = (content, digest)
2166
class InterKnit(InterVersionedFile):
2167
"""Optimised code paths for knit to knit operations."""
2169
_matching_file_from_factory = KnitVersionedFile
2170
_matching_file_to_factory = KnitVersionedFile
2173
def is_compatible(source, target):
2174
"""Be compatible with knits. """
2176
return (isinstance(source, KnitVersionedFile) and
2177
isinstance(target, KnitVersionedFile))
2178
except AttributeError:
2181
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2182
"""See InterVersionedFile.join."""
2183
assert isinstance(self.source, KnitVersionedFile)
2184
assert isinstance(self.target, KnitVersionedFile)
2186
# If the source and target are mismatched w.r.t. annotations vs
2187
# plain, the data needs to be converted accordingly
2188
if self.source.factory.annotated == self.target.factory.annotated:
2190
elif self.source.factory.annotated:
2191
converter = self._anno_to_plain_converter
2193
# We're converting from a plain to an annotated knit. This requires
2194
# building the annotations from scratch. The generic join code
2195
# handles this implicitly so we delegate to it.
2196
return super(InterKnit, self).join(pb, msg, version_ids,
2199
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2203
pb = ui.ui_factory.nested_progress_bar()
2205
version_ids = list(version_ids)
2206
if None in version_ids:
2207
version_ids.remove(None)
2209
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2210
this_versions = set(self.target._index.get_versions())
2211
# XXX: For efficiency we should not look at the whole index,
2212
# we only need to consider the referenced revisions - they
2213
# must all be present, or the method must be full-text.
2214
# TODO, RBC 20070919
2215
needed_versions = self.source_ancestry - this_versions
2217
if not needed_versions:
2219
full_list = topo_sort(self.source.get_graph())
2221
version_list = [i for i in full_list if (not self.target.has_version(i)
2222
and i in needed_versions)]
2226
copy_queue_records = []
2228
for version_id in version_list:
2229
options = self.source._index.get_options(version_id)
2230
parents = self.source._index.get_parents_with_ghosts(version_id)
2231
# check that its will be a consistent copy:
2232
for parent in parents:
2233
# if source has the parent, we must :
2234
# * already have it or
2235
# * have it scheduled already
2236
# otherwise we don't care
2237
assert (self.target.has_version(parent) or
2238
parent in copy_set or
2239
not self.source.has_version(parent))
2240
index_memo = self.source._index.get_position(version_id)
2241
copy_queue_records.append((version_id, index_memo))
2242
copy_queue.append((version_id, options, parents))
2243
copy_set.add(version_id)
2245
# data suck the join:
2247
total = len(version_list)
2250
for (version_id, raw_data), \
2251
(version_id2, options, parents) in \
2252
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2254
assert version_id == version_id2, 'logic error, inconsistent results'
2256
pb.update("Joining knit", count, total)
2258
size, raw_data = converter(raw_data, version_id, options,
2261
size = len(raw_data)
2262
raw_records.append((version_id, options, parents, size))
2263
raw_datum.append(raw_data)
2264
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2269
def _anno_to_plain_converter(self, raw_data, version_id, options,
2271
"""Convert annotated content to plain content."""
2272
data, digest = self.source._data._parse_record(version_id, raw_data)
2273
if 'fulltext' in options:
2274
content = self.source.factory.parse_fulltext(data, version_id)
2275
lines = self.target.factory.lower_fulltext(content)
2277
delta = self.source.factory.parse_line_delta(data, version_id,
2279
lines = self.target.factory.lower_line_delta(delta)
2280
return self.target._data._record_to_data(version_id, digest, lines)
2283
InterVersionedFile.register_optimiser(InterKnit)
2286
class WeaveToKnit(InterVersionedFile):
2287
"""Optimised code paths for weave to knit operations."""
2289
_matching_file_from_factory = bzrlib.weave.WeaveFile
2290
_matching_file_to_factory = KnitVersionedFile
2293
def is_compatible(source, target):
2294
"""Be compatible with weaves to knits."""
2296
return (isinstance(source, bzrlib.weave.Weave) and
2297
isinstance(target, KnitVersionedFile))
2298
except AttributeError:
2301
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2302
"""See InterVersionedFile.join."""
2303
assert isinstance(self.source, bzrlib.weave.Weave)
2304
assert isinstance(self.target, KnitVersionedFile)
2306
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2311
pb = ui.ui_factory.nested_progress_bar()
2313
version_ids = list(version_ids)
2315
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2316
this_versions = set(self.target._index.get_versions())
2317
needed_versions = self.source_ancestry - this_versions
2319
if not needed_versions:
2321
full_list = topo_sort(self.source.get_graph())
2323
version_list = [i for i in full_list if (not self.target.has_version(i)
2324
and i in needed_versions)]
2328
total = len(version_list)
2329
for version_id in version_list:
2330
pb.update("Converting to knit", count, total)
2331
parents = self.source.get_parents(version_id)
2332
# check that its will be a consistent copy:
2333
for parent in parents:
2334
# if source has the parent, we must already have it
2335
assert (self.target.has_version(parent))
2336
self.target.add_lines(
2337
version_id, parents, self.source.get_lines(version_id))
2344
InterVersionedFile.register_optimiser(WeaveToKnit)
2347
# Deprecated, use PatienceSequenceMatcher instead
2348
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2351
def annotate_knit(knit, revision_id):
2352
"""Annotate a knit with no cached annotations.
2354
This implementation is for knits with no cached annotations.
2355
It will work for knits with cached annotations, but this is not
2358
ancestry = knit.get_ancestry(revision_id)
2359
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2361
for candidate in ancestry:
2362
if candidate in annotations:
2364
parents = knit.get_parents(candidate)
2365
if len(parents) == 0:
2367
elif knit._index.get_method(candidate) != 'line-delta':
2370
parent, sha1, noeol, delta = knit.get_delta(candidate)
2371
blocks = KnitContent.get_line_delta_blocks(delta,
2372
fulltext[parents[0]], fulltext[candidate])
2373
annotations[candidate] = list(annotate.reannotate([annotations[p]
2374
for p in parents], fulltext[candidate], candidate, blocks))
2375
return iter(annotations[revision_id])
2379
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2381
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
