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
104
from bzrlib.osutils import (
109
from bzrlib.symbol_versioning import DEPRECATED_PARAMETER, deprecated_passed
110
from bzrlib.tsort import topo_sort
113
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
116
# TODO: Split out code specific to this format into an associated object.
118
# TODO: Can we put in some kind of value to check that the index and data
119
# files belong together?
121
# TODO: accommodate binaries, perhaps by storing a byte count
123
# TODO: function to check whole file
125
# TODO: atomically append data, then measure backwards from the cursor
126
# position after writing to work out where it was located. we may need to
127
# bypass python file buffering.
129
DATA_SUFFIX = '.knit'
130
INDEX_SUFFIX = '.kndx'
133
class KnitContent(object):
134
"""Content of a knit version to which deltas can be applied."""
137
"""Return a list of (origin, text) tuples."""
138
return list(self.annotate_iter())
140
def line_delta_iter(self, new_lines):
141
"""Generate line-based delta from this content to new_lines."""
142
new_texts = new_lines.text()
143
old_texts = self.text()
144
s = patiencediff.PatienceSequenceMatcher(None, old_texts, new_texts)
145
for tag, i1, i2, j1, j2 in s.get_opcodes():
148
# ofrom, oto, length, data
149
yield i1, i2, j2 - j1, new_lines._lines[j1:j2]
151
def line_delta(self, new_lines):
152
return list(self.line_delta_iter(new_lines))
155
def get_line_delta_blocks(knit_delta, source, target):
156
"""Extract SequenceMatcher.get_matching_blocks() from a knit delta"""
157
target_len = len(target)
160
for s_begin, s_end, t_len, new_text in knit_delta:
161
true_n = s_begin - s_pos
164
# knit deltas do not provide reliable info about whether the
165
# last line of a file matches, due to eol handling.
166
if source[s_pos + n -1] != target[t_pos + n -1]:
169
yield s_pos, t_pos, n
170
t_pos += t_len + true_n
172
n = target_len - t_pos
174
if source[s_pos + n -1] != target[t_pos + n -1]:
177
yield s_pos, t_pos, n
178
yield s_pos + (target_len - t_pos), target_len, 0
181
class AnnotatedKnitContent(KnitContent):
182
"""Annotated content."""
184
def __init__(self, lines):
187
def annotate_iter(self):
188
"""Yield tuples of (origin, text) for each content line."""
189
return iter(self._lines)
191
def strip_last_line_newline(self):
192
line = self._lines[-1][1].rstrip('\n')
193
self._lines[-1] = (self._lines[-1][0], line)
196
return [text for origin, text in self._lines]
199
return AnnotatedKnitContent(self._lines[:])
202
class PlainKnitContent(KnitContent):
203
"""Unannotated content.
205
When annotate[_iter] is called on this content, the same version is reported
206
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
210
def __init__(self, lines, version_id):
212
self._version_id = version_id
214
def annotate_iter(self):
215
"""Yield tuples of (origin, text) for each content line."""
216
for line in self._lines:
217
yield self._version_id, line
220
return PlainKnitContent(self._lines[:], self._version_id)
222
def strip_last_line_newline(self):
223
self._lines[-1] = self._lines[-1].rstrip('\n')
229
class KnitAnnotateFactory(object):
230
"""Factory for creating annotated Content objects."""
234
def make(self, lines, version_id):
235
num_lines = len(lines)
236
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
238
def parse_fulltext(self, content, version_id):
239
"""Convert fulltext to internal representation
241
fulltext content is of the format
242
revid(utf8) plaintext\n
243
internal representation is of the format:
246
# TODO: jam 20070209 The tests expect this to be returned as tuples,
247
# but the code itself doesn't really depend on that.
248
# Figure out a way to not require the overhead of turning the
249
# list back into tuples.
250
lines = [tuple(line.split(' ', 1)) for line in content]
251
return AnnotatedKnitContent(lines)
253
def parse_line_delta_iter(self, lines):
254
return iter(self.parse_line_delta(lines))
256
def parse_line_delta(self, lines, version_id):
257
"""Convert a line based delta into internal representation.
259
line delta is in the form of:
260
intstart intend intcount
262
revid(utf8) newline\n
263
internal representation is
264
(start, end, count, [1..count tuples (revid, newline)])
271
def cache_and_return(line):
272
origin, text = line.split(' ', 1)
273
return cache.setdefault(origin, origin), text
275
# walk through the lines parsing.
277
start, end, count = [int(n) for n in header.split(',')]
278
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
279
result.append((start, end, count, contents))
282
def get_fulltext_content(self, lines):
283
"""Extract just the content lines from a fulltext."""
284
return (line.split(' ', 1)[1] for line in lines)
286
def get_linedelta_content(self, lines):
287
"""Extract just the content from a line delta.
289
This doesn't return all of the extra information stored in a delta.
290
Only the actual content lines.
295
header = header.split(',')
296
count = int(header[2])
297
for i in xrange(count):
298
origin, text = next().split(' ', 1)
301
def lower_fulltext(self, content):
302
"""convert a fulltext content record into a serializable form.
304
see parse_fulltext which this inverts.
306
# TODO: jam 20070209 We only do the caching thing to make sure that
307
# the origin is a valid utf-8 line, eventually we could remove it
308
return ['%s %s' % (o, t) for o, t in content._lines]
310
def lower_line_delta(self, delta):
311
"""convert a delta into a serializable form.
313
See parse_line_delta which this inverts.
315
# TODO: jam 20070209 We only do the caching thing to make sure that
316
# the origin is a valid utf-8 line, eventually we could remove it
318
for start, end, c, lines in delta:
319
out.append('%d,%d,%d\n' % (start, end, c))
320
out.extend(origin + ' ' + text
321
for origin, text in lines)
324
def annotate_iter(self, knit, version_id):
325
content = knit._get_content(version_id)
326
return content.annotate_iter()
329
class KnitPlainFactory(object):
330
"""Factory for creating plain Content objects."""
334
def make(self, lines, version_id):
335
return PlainKnitContent(lines, version_id)
337
def parse_fulltext(self, content, version_id):
338
"""This parses an unannotated fulltext.
340
Note that this is not a noop - the internal representation
341
has (versionid, line) - its just a constant versionid.
343
return self.make(content, version_id)
345
def parse_line_delta_iter(self, lines, version_id):
347
num_lines = len(lines)
348
while cur < num_lines:
351
start, end, c = [int(n) for n in header.split(',')]
352
yield start, end, c, lines[cur:cur+c]
355
def parse_line_delta(self, lines, version_id):
356
return list(self.parse_line_delta_iter(lines, version_id))
358
def get_fulltext_content(self, lines):
359
"""Extract just the content lines from a fulltext."""
362
def get_linedelta_content(self, lines):
363
"""Extract just the content from a line delta.
365
This doesn't return all of the extra information stored in a delta.
366
Only the actual content lines.
371
header = header.split(',')
372
count = int(header[2])
373
for i in xrange(count):
376
def lower_fulltext(self, content):
377
return content.text()
379
def lower_line_delta(self, delta):
381
for start, end, c, lines in delta:
382
out.append('%d,%d,%d\n' % (start, end, c))
386
def annotate_iter(self, knit, version_id):
387
return annotate_knit(knit, version_id)
390
def make_empty_knit(transport, relpath):
391
"""Construct a empty knit at the specified location."""
392
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
395
class KnitVersionedFile(VersionedFile):
396
"""Weave-like structure with faster random access.
398
A knit stores a number of texts and a summary of the relationships
399
between them. Texts are identified by a string version-id. Texts
400
are normally stored and retrieved as a series of lines, but can
401
also be passed as single strings.
403
Lines are stored with the trailing newline (if any) included, to
404
avoid special cases for files with no final newline. Lines are
405
composed of 8-bit characters, not unicode. The combination of
406
these approaches should mean any 'binary' file can be safely
407
stored and retrieved.
410
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
411
factory=None, basis_knit=DEPRECATED_PARAMETER, delta=True,
412
create=False, create_parent_dir=False, delay_create=False,
413
dir_mode=None, index=None, access_method=None):
414
"""Construct a knit at location specified by relpath.
416
:param create: If not True, only open an existing knit.
417
:param create_parent_dir: If True, create the parent directory if
418
creating the file fails. (This is used for stores with
419
hash-prefixes that may not exist yet)
420
:param delay_create: The calling code is aware that the knit won't
421
actually be created until the first data is stored.
422
:param index: An index to use for the knit.
424
if deprecated_passed(basis_knit):
425
warnings.warn("KnitVersionedFile.__(): The basis_knit parameter is"
426
" deprecated as of bzr 0.9.",
427
DeprecationWarning, stacklevel=2)
428
if access_mode is None:
430
super(KnitVersionedFile, self).__init__(access_mode)
431
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
432
self.transport = transport
433
self.filename = relpath
434
self.factory = factory or KnitAnnotateFactory()
435
self.writable = (access_mode == 'w')
438
self._max_delta_chain = 200
441
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
442
access_mode, create=create, file_mode=file_mode,
443
create_parent_dir=create_parent_dir, delay_create=delay_create,
447
if access_method is None:
448
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
449
((create and not len(self)) and delay_create), create_parent_dir)
451
_access = access_method
452
if create and not len(self) and not delay_create:
454
self._data = _KnitData(_access)
457
return '%s(%s)' % (self.__class__.__name__,
458
self.transport.abspath(self.filename))
460
def _check_should_delta(self, first_parents):
461
"""Iterate back through the parent listing, looking for a fulltext.
463
This is used when we want to decide whether to add a delta or a new
464
fulltext. It searches for _max_delta_chain parents. When it finds a
465
fulltext parent, it sees if the total size of the deltas leading up to
466
it is large enough to indicate that we want a new full text anyway.
468
Return True if we should create a new delta, False if we should use a
473
delta_parents = first_parents
474
for count in xrange(self._max_delta_chain):
475
parent = delta_parents[0]
476
method = self._index.get_method(parent)
477
index, pos, size = self._index.get_position(parent)
478
if method == 'fulltext':
482
delta_parents = self._index.get_parents(parent)
484
# We couldn't find a fulltext, so we must create a new one
487
return fulltext_size > delta_size
489
def _add_raw_records(self, records, data):
490
"""Add all the records 'records' with data pre-joined in 'data'.
492
:param records: A list of tuples(version_id, options, parents, size).
493
:param data: The data for the records. When it is written, the records
494
are adjusted to have pos pointing into data by the sum of
495
the preceding records sizes.
498
raw_record_sizes = [record[3] for record in records]
499
positions = self._data.add_raw_records(raw_record_sizes, data)
502
for (version_id, options, parents, size), access_memo in zip(
504
index_entries.append((version_id, options, access_memo, parents))
505
if self._data._do_cache:
506
self._data._cache[version_id] = data[offset:offset+size]
508
self._index.add_versions(index_entries)
510
def enable_cache(self):
511
"""Start caching data for this knit"""
512
self._data.enable_cache()
514
def clear_cache(self):
515
"""Clear the data cache only."""
516
self._data.clear_cache()
518
def copy_to(self, name, transport):
519
"""See VersionedFile.copy_to()."""
520
# copy the current index to a temp index to avoid racing with local
522
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
523
self.transport.get(self._index._filename))
525
f = self._data._open_file()
527
transport.put_file(name + DATA_SUFFIX, f)
530
# move the copied index into place
531
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
533
def create_empty(self, name, transport, mode=None):
534
return KnitVersionedFile(name, transport, factory=self.factory,
535
delta=self.delta, create=True)
537
def get_data_stream(self, required_versions):
538
"""Get a data stream for the specified versions.
540
Versions may be returned in any order, not necessarily the order
543
:param required_versions: The exact set of versions to be extracted.
544
Unlike some other knit methods, this is not used to generate a
545
transitive closure, rather it is used precisely as given.
547
:returns: format_signature, list of (version, options, length, parents),
550
required_versions = set([osutils.safe_revision_id(v) for v in
552
# we don't care about inclusions, the caller cares.
553
# but we need to setup a list of records to visit.
554
for version_id in required_versions:
555
if not self.has_version(version_id):
556
raise RevisionNotPresent(version_id, self.filename)
557
# Pick the desired versions out of the index in oldest-to-newest order
559
for version_id in self.versions():
560
if version_id in required_versions:
561
version_list.append(version_id)
563
# create the list of version information for the result
564
copy_queue_records = []
566
result_version_list = []
567
for version_id in version_list:
568
options = self._index.get_options(version_id)
569
parents = self._index.get_parents_with_ghosts(version_id)
570
index_memo = self._index.get_position(version_id)
571
copy_queue_records.append((version_id, index_memo))
572
none, data_pos, data_size = index_memo
573
copy_set.add(version_id)
574
# version, options, length, parents
575
result_version_list.append((version_id, options, data_size,
578
# Read the compressed record data.
580
# From here down to the return should really be logic in the returned
581
# callable -- in a class that adapts read_records_iter_raw to read
584
for (version_id, raw_data), \
585
(version_id2, options, _, parents) in \
586
izip(self._data.read_records_iter_raw(copy_queue_records),
587
result_version_list):
588
assert version_id == version_id2, 'logic error, inconsistent results'
589
raw_datum.append(raw_data)
590
pseudo_file = StringIO(''.join(raw_datum))
593
return pseudo_file.read()
595
return pseudo_file.read(length)
596
return (self.get_format_signature(), result_version_list, read)
598
def _extract_blocks(self, version_id, source, target):
599
if self._index.get_method(version_id) != 'line-delta':
601
parent, sha1, noeol, delta = self.get_delta(version_id)
602
return KnitContent.get_line_delta_blocks(delta, source, target)
604
def get_delta(self, version_id):
605
"""Get a delta for constructing version from some other version."""
606
version_id = osutils.safe_revision_id(version_id)
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
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
650
record_map = self._get_record_map(version_ids)
651
# record entry 2 is the 'digest'.
652
return [record_map[v][2] for v in version_ids]
656
"""See VersionedFile.get_suffixes()."""
657
return [DATA_SUFFIX, INDEX_SUFFIX]
659
def has_ghost(self, version_id):
660
"""True if there is a ghost reference in the file to version_id."""
661
version_id = osutils.safe_revision_id(version_id)
663
if self.has_version(version_id):
665
# optimisable if needed by memoising the _ghosts set.
666
items = self._index.get_graph()
667
for node, parents in items:
668
for parent in parents:
669
if parent not in self._index._cache:
670
if parent == version_id:
674
def insert_data_stream(self, (format, data_list, reader_callable)):
675
"""Insert knit records from a data stream into this knit.
677
If a version in the stream is already present in this knit, it will not
678
be inserted a second time. It will be checked for consistency with the
679
stored version however, and may cause a KnitCorrupt error to be raised
680
if the data in the stream disagrees with the already stored data.
682
:seealso: get_data_stream
684
if format != self.get_format_signature():
685
trace.mutter('incompatible format signature inserting to %r', self)
686
raise KnitDataStreamIncompatible(
687
format, self.get_format_signature())
689
for version_id, options, length, parents in data_list:
690
if self.has_version(version_id):
691
# First check: the list of parents.
692
my_parents = self.get_parents_with_ghosts(version_id)
693
if my_parents != parents:
694
# XXX: KnitCorrupt is not quite the right exception here.
697
'parents list %r from data stream does not match '
698
'already recorded parents %r for %s'
699
% (parents, my_parents, version_id))
701
# Also check the SHA-1 of the fulltext this content will
703
raw_data = reader_callable(length)
704
my_fulltext_sha1 = self.get_sha1(version_id)
705
df, rec = self._data._parse_record_header(version_id, raw_data)
706
stream_fulltext_sha1 = rec[3]
707
if my_fulltext_sha1 != stream_fulltext_sha1:
708
# Actually, we don't know if it's this knit that's corrupt,
709
# or the data stream we're trying to insert.
711
self.filename, 'sha-1 does not match %s' % version_id)
713
self._add_raw_records(
714
[(version_id, options, parents, length)],
715
reader_callable(length))
718
"""See VersionedFile.versions."""
719
if 'evil' in debug.debug_flags:
720
trace.mutter_callsite(2, "versions scales with size of history")
721
return self._index.get_versions()
723
def has_version(self, version_id):
724
"""See VersionedFile.has_version."""
725
if 'evil' in debug.debug_flags:
726
trace.mutter_callsite(2, "has_version is a LBYL scenario")
727
version_id = osutils.safe_revision_id(version_id)
728
return self._index.has_version(version_id)
730
__contains__ = has_version
732
def _merge_annotations(self, content, parents, parent_texts={},
733
delta=None, annotated=None,
734
left_matching_blocks=None):
735
"""Merge annotations for content. This is done by comparing
736
the annotations based on changed to the text.
738
if left_matching_blocks is not None:
739
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
743
for parent_id in parents:
744
merge_content = self._get_content(parent_id, parent_texts)
745
if (parent_id == parents[0] and delta_seq is not None):
748
seq = patiencediff.PatienceSequenceMatcher(
749
None, merge_content.text(), content.text())
750
for i, j, n in seq.get_matching_blocks():
753
# this appears to copy (origin, text) pairs across to the
754
# new content for any line that matches the last-checked
756
content._lines[j:j+n] = merge_content._lines[i:i+n]
758
if delta_seq is None:
759
reference_content = self._get_content(parents[0], parent_texts)
760
new_texts = content.text()
761
old_texts = reference_content.text()
762
delta_seq = patiencediff.PatienceSequenceMatcher(
763
None, old_texts, new_texts)
764
return self._make_line_delta(delta_seq, content)
766
def _make_line_delta(self, delta_seq, new_content):
767
"""Generate a line delta from delta_seq and new_content."""
769
for op in delta_seq.get_opcodes():
772
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
775
def _get_components_positions(self, version_ids):
776
"""Produce a map of position data for the components of versions.
778
This data is intended to be used for retrieving the knit records.
780
A dict of version_id to (method, data_pos, data_size, next) is
782
method is the way referenced data should be applied.
783
data_pos is the position of the data in the knit.
784
data_size is the size of the data in the knit.
785
next is the build-parent of the version, or None for fulltexts.
788
for version_id in version_ids:
791
while cursor is not None and cursor not in component_data:
792
method = self._index.get_method(cursor)
793
if method == 'fulltext':
796
next = self.get_parents(cursor)[0]
797
index_memo = self._index.get_position(cursor)
798
component_data[cursor] = (method, index_memo, next)
800
return component_data
802
def _get_content(self, version_id, parent_texts={}):
803
"""Returns a content object that makes up the specified
805
cached_version = parent_texts.get(version_id, None)
806
if cached_version is not None:
807
if not self.has_version(version_id):
808
raise RevisionNotPresent(version_id, self.filename)
809
return cached_version
811
text_map, contents_map = self._get_content_maps([version_id])
812
return contents_map[version_id]
814
def _check_versions_present(self, version_ids):
815
"""Check that all specified versions are present."""
816
self._index.check_versions_present(version_ids)
818
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
819
nostore_sha, random_id, check_content):
820
"""See VersionedFile.add_lines_with_ghosts()."""
821
self._check_add(version_id, lines, random_id, check_content)
822
return self._add(version_id, lines, parents, self.delta,
823
parent_texts, None, nostore_sha)
825
def _add_lines(self, version_id, parents, lines, parent_texts,
826
left_matching_blocks, nostore_sha, random_id, check_content):
827
"""See VersionedFile.add_lines."""
828
self._check_add(version_id, lines, random_id, check_content)
829
self._check_versions_present(parents)
830
return self._add(version_id, lines[:], parents, self.delta,
831
parent_texts, left_matching_blocks, nostore_sha)
833
def _check_add(self, version_id, lines, random_id, check_content):
834
"""check that version_id and lines are safe to add."""
835
if contains_whitespace(version_id):
836
raise InvalidRevisionId(version_id, self.filename)
837
self.check_not_reserved_id(version_id)
838
# Technically this could be avoided if we are happy to allow duplicate
839
# id insertion when other things than bzr core insert texts, but it
840
# seems useful for folk using the knit api directly to have some safety
841
# blanket that we can disable.
842
if not random_id and self.has_version(version_id):
843
raise RevisionAlreadyPresent(version_id, self.filename)
845
self._check_lines_not_unicode(lines)
846
self._check_lines_are_lines(lines)
848
def _add(self, version_id, lines, parents, delta, parent_texts,
849
left_matching_blocks, nostore_sha):
850
"""Add a set of lines on top of version specified by parents.
852
If delta is true, compress the text as a line-delta against
855
Any versions not present will be converted into ghosts.
857
# 461 0 6546.0390 43.9100 bzrlib.knit:489(_add)
858
# +400 0 889.4890 418.9790 +bzrlib.knit:192(lower_fulltext)
859
# +461 0 1364.8070 108.8030 +bzrlib.knit:996(add_record)
860
# +461 0 193.3940 41.5720 +bzrlib.knit:898(add_version)
861
# +461 0 134.0590 18.3810 +bzrlib.osutils:361(sha_strings)
862
# +461 0 36.3420 15.4540 +bzrlib.knit:146(make)
863
# +1383 0 8.0370 8.0370 +<len>
864
# +61 0 13.5770 7.9190 +bzrlib.knit:199(lower_line_delta)
865
# +61 0 963.3470 7.8740 +bzrlib.knit:427(_get_content)
866
# +61 0 973.9950 5.2950 +bzrlib.knit:136(line_delta)
867
# +61 0 1918.1800 5.2640 +bzrlib.knit:359(_merge_annotations)
870
if parent_texts is None:
872
for parent in parents:
873
if self.has_version(parent):
874
present_parents.append(parent)
876
# can only compress against the left most present parent.
878
(len(present_parents) == 0 or
879
present_parents[0] != parents[0])):
882
digest = sha_strings(lines)
883
if nostore_sha == digest:
884
raise errors.ExistingContent
885
text_length = sum(map(len, lines))
888
if lines[-1][-1] != '\n':
889
# copy the contents of lines.
891
options.append('no-eol')
892
lines[-1] = lines[-1] + '\n'
895
# To speed the extract of texts the delta chain is limited
896
# to a fixed number of deltas. This should minimize both
897
# I/O and the time spend applying deltas.
898
delta = self._check_should_delta(present_parents)
900
assert isinstance(version_id, str)
901
content = self.factory.make(lines, version_id)
902
if delta or (self.factory.annotated and len(present_parents) > 0):
903
# Merge annotations from parent texts if needed.
904
delta_hunks = self._merge_annotations(content, present_parents,
905
parent_texts, delta, self.factory.annotated,
906
left_matching_blocks)
909
options.append('line-delta')
910
store_lines = self.factory.lower_line_delta(delta_hunks)
912
options.append('fulltext')
913
store_lines = self.factory.lower_fulltext(content)
915
access_memo = self._data.add_record(version_id, digest, store_lines)
916
self._index.add_version(version_id, options, access_memo, parents)
917
return digest, text_length, content
919
def check(self, progress_bar=None):
920
"""See VersionedFile.check()."""
922
def _clone_text(self, new_version_id, old_version_id, parents):
923
"""See VersionedFile.clone_text()."""
924
# FIXME RBC 20060228 make fast by only inserting an index with null
926
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
928
def get_lines(self, version_id):
929
"""See VersionedFile.get_lines()."""
930
return self.get_line_list([version_id])[0]
932
def _get_record_map(self, version_ids):
933
"""Produce a dictionary of knit records.
935
The keys are version_ids, the values are tuples of (method, content,
937
method is the way the content should be applied.
938
content is a KnitContent object.
939
digest is the SHA1 digest of this version id after all steps are done
940
next is the build-parent of the version, i.e. the leftmost ancestor.
941
If the method is fulltext, next will be None.
943
position_map = self._get_components_positions(version_ids)
944
# c = component_id, m = method, i_m = index_memo, n = next
945
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
947
for component_id, content, digest in \
948
self._data.read_records_iter(records):
949
method, index_memo, next = position_map[component_id]
950
record_map[component_id] = method, content, digest, next
954
def get_text(self, version_id):
955
"""See VersionedFile.get_text"""
956
return self.get_texts([version_id])[0]
958
def get_texts(self, version_ids):
959
return [''.join(l) for l in self.get_line_list(version_ids)]
961
def get_line_list(self, version_ids):
962
"""Return the texts of listed versions as a list of strings."""
963
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
964
for version_id in version_ids:
965
self.check_not_reserved_id(version_id)
966
text_map, content_map = self._get_content_maps(version_ids)
967
return [text_map[v] for v in version_ids]
969
_get_lf_split_line_list = get_line_list
971
def _get_content_maps(self, version_ids):
972
"""Produce maps of text and KnitContents
974
:return: (text_map, content_map) where text_map contains the texts for
975
the requested versions and content_map contains the KnitContents.
976
Both dicts take version_ids as their keys.
978
for version_id in version_ids:
979
if not self.has_version(version_id):
980
raise RevisionNotPresent(version_id, self.filename)
981
record_map = self._get_record_map(version_ids)
986
for version_id in version_ids:
989
while cursor is not None:
990
method, data, digest, next = record_map[cursor]
991
components.append((cursor, method, data, digest))
992
if cursor in content_map:
997
for component_id, method, data, digest in reversed(components):
998
if component_id in content_map:
999
content = content_map[component_id]
1001
if method == 'fulltext':
1002
assert content is None
1003
content = self.factory.parse_fulltext(data, version_id)
1004
elif method == 'line-delta':
1005
delta = self.factory.parse_line_delta(data, version_id)
1006
content = content.copy()
1007
content._lines = self._apply_delta(content._lines,
1009
content_map[component_id] = content
1011
if 'no-eol' in self._index.get_options(version_id):
1012
content = content.copy()
1013
content.strip_last_line_newline()
1014
final_content[version_id] = content
1016
# digest here is the digest from the last applied component.
1017
text = content.text()
1018
if sha_strings(text) != digest:
1019
raise KnitCorrupt(self.filename,
1020
'sha-1 does not match %s' % version_id)
1022
text_map[version_id] = text
1023
return text_map, final_content
1026
def _apply_delta(lines, delta):
1027
"""Apply delta to lines."""
1030
for start, end, count, delta_lines in delta:
1031
lines[offset+start:offset+end] = delta_lines
1032
offset = offset + (start - end) + count
1035
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1037
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1038
if version_ids is None:
1039
version_ids = self.versions()
1041
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
1043
pb = progress.DummyProgress()
1044
# we don't care about inclusions, the caller cares.
1045
# but we need to setup a list of records to visit.
1046
# we need version_id, position, length
1047
version_id_records = []
1048
requested_versions = set(version_ids)
1049
# filter for available versions
1050
for version_id in requested_versions:
1051
if not self.has_version(version_id):
1052
raise RevisionNotPresent(version_id, self.filename)
1053
# get a in-component-order queue:
1054
for version_id in self.versions():
1055
if version_id in requested_versions:
1056
index_memo = self._index.get_position(version_id)
1057
version_id_records.append((version_id, index_memo))
1059
total = len(version_id_records)
1060
for version_idx, (version_id, data, sha_value) in \
1061
enumerate(self._data.read_records_iter(version_id_records)):
1062
pb.update('Walking content.', version_idx, total)
1063
method = self._index.get_method(version_id)
1065
assert method in ('fulltext', 'line-delta')
1066
if method == 'fulltext':
1067
line_iterator = self.factory.get_fulltext_content(data)
1069
line_iterator = self.factory.get_linedelta_content(data)
1070
for line in line_iterator:
1073
pb.update('Walking content.', total, total)
1075
def iter_parents(self, version_ids):
1076
"""Iterate through the parents for many version ids.
1078
:param version_ids: An iterable yielding version_ids.
1079
:return: An iterator that yields (version_id, parents). Requested
1080
version_ids not present in the versioned file are simply skipped.
1081
The order is undefined, allowing for different optimisations in
1082
the underlying implementation.
1084
version_ids = [osutils.safe_revision_id(version_id) for
1085
version_id in version_ids]
1086
return self._index.iter_parents(version_ids)
1088
def num_versions(self):
1089
"""See VersionedFile.num_versions()."""
1090
return self._index.num_versions()
1092
__len__ = num_versions
1094
def annotate_iter(self, version_id):
1095
"""See VersionedFile.annotate_iter."""
1096
version_id = osutils.safe_revision_id(version_id)
1097
return self.factory.annotate_iter(self, version_id)
1099
def get_parents(self, version_id):
1100
"""See VersionedFile.get_parents."""
1103
# 52554 calls in 1264 872 internal down from 3674
1104
version_id = osutils.safe_revision_id(version_id)
1106
return self._index.get_parents(version_id)
1108
raise RevisionNotPresent(version_id, self.filename)
1110
def get_parents_with_ghosts(self, version_id):
1111
"""See VersionedFile.get_parents."""
1112
version_id = osutils.safe_revision_id(version_id)
1114
return self._index.get_parents_with_ghosts(version_id)
1116
raise RevisionNotPresent(version_id, self.filename)
1118
def get_ancestry(self, versions, topo_sorted=True):
1119
"""See VersionedFile.get_ancestry."""
1120
if isinstance(versions, basestring):
1121
versions = [versions]
1124
versions = [osutils.safe_revision_id(v) for v in versions]
1125
return self._index.get_ancestry(versions, topo_sorted)
1127
def get_ancestry_with_ghosts(self, versions):
1128
"""See VersionedFile.get_ancestry_with_ghosts."""
1129
if isinstance(versions, basestring):
1130
versions = [versions]
1133
versions = [osutils.safe_revision_id(v) for v in versions]
1134
return self._index.get_ancestry_with_ghosts(versions)
1136
def plan_merge(self, ver_a, ver_b):
1137
"""See VersionedFile.plan_merge."""
1138
ver_a = osutils.safe_revision_id(ver_a)
1139
ver_b = osutils.safe_revision_id(ver_b)
1140
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1142
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1143
annotated_a = self.annotate(ver_a)
1144
annotated_b = self.annotate(ver_b)
1145
return merge._plan_annotate_merge(annotated_a, annotated_b,
1146
ancestors_a, ancestors_b)
1149
class _KnitComponentFile(object):
1150
"""One of the files used to implement a knit database"""
1152
def __init__(self, transport, filename, mode, file_mode=None,
1153
create_parent_dir=False, dir_mode=None):
1154
self._transport = transport
1155
self._filename = filename
1157
self._file_mode = file_mode
1158
self._dir_mode = dir_mode
1159
self._create_parent_dir = create_parent_dir
1160
self._need_to_create = False
1162
def _full_path(self):
1163
"""Return the full path to this file."""
1164
return self._transport.base + self._filename
1166
def check_header(self, fp):
1167
line = fp.readline()
1169
# An empty file can actually be treated as though the file doesn't
1171
raise errors.NoSuchFile(self._full_path())
1172
if line != self.HEADER:
1173
raise KnitHeaderError(badline=line,
1174
filename=self._transport.abspath(self._filename))
1177
return '%s(%s)' % (self.__class__.__name__, self._filename)
1180
class _KnitIndex(_KnitComponentFile):
1181
"""Manages knit index file.
1183
The index is already kept in memory and read on startup, to enable
1184
fast lookups of revision information. The cursor of the index
1185
file is always pointing to the end, making it easy to append
1188
_cache is a cache for fast mapping from version id to a Index
1191
_history is a cache for fast mapping from indexes to version ids.
1193
The index data format is dictionary compressed when it comes to
1194
parent references; a index entry may only have parents that with a
1195
lover index number. As a result, the index is topological sorted.
1197
Duplicate entries may be written to the index for a single version id
1198
if this is done then the latter one completely replaces the former:
1199
this allows updates to correct version and parent information.
1200
Note that the two entries may share the delta, and that successive
1201
annotations and references MUST point to the first entry.
1203
The index file on disc contains a header, followed by one line per knit
1204
record. The same revision can be present in an index file more than once.
1205
The first occurrence gets assigned a sequence number starting from 0.
1207
The format of a single line is
1208
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1209
REVISION_ID is a utf8-encoded revision id
1210
FLAGS is a comma separated list of flags about the record. Values include
1211
no-eol, line-delta, fulltext.
1212
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1213
that the the compressed data starts at.
1214
LENGTH is the ascii representation of the length of the data file.
1215
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1217
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1218
revision id already in the knit that is a parent of REVISION_ID.
1219
The ' :' marker is the end of record marker.
1222
when a write is interrupted to the index file, it will result in a line
1223
that does not end in ' :'. If the ' :' is not present at the end of a line,
1224
or at the end of the file, then the record that is missing it will be
1225
ignored by the parser.
1227
When writing new records to the index file, the data is preceded by '\n'
1228
to ensure that records always start on new lines even if the last write was
1229
interrupted. As a result its normal for the last line in the index to be
1230
missing a trailing newline. One can be added with no harmful effects.
1233
HEADER = "# bzr knit index 8\n"
1235
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1236
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1238
def _cache_version(self, version_id, options, pos, size, parents):
1239
"""Cache a version record in the history array and index cache.
1241
This is inlined into _load_data for performance. KEEP IN SYNC.
1242
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1245
# only want the _history index to reference the 1st index entry
1247
if version_id not in self._cache:
1248
index = len(self._history)
1249
self._history.append(version_id)
1251
index = self._cache[version_id][5]
1252
self._cache[version_id] = (version_id,
1259
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1260
create_parent_dir=False, delay_create=False, dir_mode=None):
1261
_KnitComponentFile.__init__(self, transport, filename, mode,
1262
file_mode=file_mode,
1263
create_parent_dir=create_parent_dir,
1266
# position in _history is the 'official' index for a revision
1267
# but the values may have come from a newer entry.
1268
# so - wc -l of a knit index is != the number of unique names
1272
fp = self._transport.get(self._filename)
1274
# _load_data may raise NoSuchFile if the target knit is
1276
_load_data(self, fp)
1280
if mode != 'w' or not create:
1283
self._need_to_create = True
1285
self._transport.put_bytes_non_atomic(
1286
self._filename, self.HEADER, mode=self._file_mode)
1288
def get_graph(self):
1289
"""Return a list of the node:parents lists from this knit index."""
1290
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1292
def get_ancestry(self, versions, topo_sorted=True):
1293
"""See VersionedFile.get_ancestry."""
1294
# get a graph of all the mentioned versions:
1296
pending = set(versions)
1299
version = pending.pop()
1302
parents = [p for p in cache[version][4] if p in cache]
1304
raise RevisionNotPresent(version, self._filename)
1305
# if not completed and not a ghost
1306
pending.update([p for p in parents if p not in graph])
1307
graph[version] = parents
1310
return topo_sort(graph.items())
1312
def get_ancestry_with_ghosts(self, versions):
1313
"""See VersionedFile.get_ancestry_with_ghosts."""
1314
# get a graph of all the mentioned versions:
1315
self.check_versions_present(versions)
1318
pending = set(versions)
1320
version = pending.pop()
1322
parents = cache[version][4]
1328
pending.update([p for p in parents if p not in graph])
1329
graph[version] = parents
1330
return topo_sort(graph.items())
1332
def iter_parents(self, version_ids):
1333
"""Iterate through the parents for many version ids.
1335
:param version_ids: An iterable yielding version_ids.
1336
:return: An iterator that yields (version_id, parents). Requested
1337
version_ids not present in the versioned file are simply skipped.
1338
The order is undefined, allowing for different optimisations in
1339
the underlying implementation.
1341
for version_id in version_ids:
1343
yield version_id, tuple(self.get_parents(version_id))
1347
def num_versions(self):
1348
return len(self._history)
1350
__len__ = num_versions
1352
def get_versions(self):
1353
"""Get all the versions in the file. not topologically sorted."""
1354
return self._history
1356
def _version_list_to_index(self, versions):
1359
for version in versions:
1360
if version in cache:
1361
# -- inlined lookup() --
1362
result_list.append(str(cache[version][5]))
1363
# -- end lookup () --
1365
result_list.append('.' + version)
1366
return ' '.join(result_list)
1368
def add_version(self, version_id, options, index_memo, parents):
1369
"""Add a version record to the index."""
1370
self.add_versions(((version_id, options, index_memo, parents),))
1372
def add_versions(self, versions):
1373
"""Add multiple versions to the index.
1375
:param versions: a list of tuples:
1376
(version_id, options, pos, size, parents).
1379
orig_history = self._history[:]
1380
orig_cache = self._cache.copy()
1383
for version_id, options, (index, pos, size), parents in versions:
1384
line = "\n%s %s %s %s %s :" % (version_id,
1388
self._version_list_to_index(parents))
1389
assert isinstance(line, str), \
1390
'content must be utf-8 encoded: %r' % (line,)
1392
self._cache_version(version_id, options, pos, size, parents)
1393
if not self._need_to_create:
1394
self._transport.append_bytes(self._filename, ''.join(lines))
1397
sio.write(self.HEADER)
1398
sio.writelines(lines)
1400
self._transport.put_file_non_atomic(self._filename, sio,
1401
create_parent_dir=self._create_parent_dir,
1402
mode=self._file_mode,
1403
dir_mode=self._dir_mode)
1404
self._need_to_create = False
1406
# If any problems happen, restore the original values and re-raise
1407
self._history = orig_history
1408
self._cache = orig_cache
1411
def has_version(self, version_id):
1412
"""True if the version is in the index."""
1413
return version_id in self._cache
1415
def get_position(self, version_id):
1416
"""Return details needed to access the version.
1418
.kndx indices do not support split-out data, so return None for the
1421
:return: a tuple (None, data position, size) to hand to the access
1422
logic to get the record.
1424
entry = self._cache[version_id]
1425
return None, entry[2], entry[3]
1427
def get_method(self, version_id):
1428
"""Return compression method of specified version."""
1429
options = self._cache[version_id][1]
1430
if 'fulltext' in options:
1433
if 'line-delta' not in options:
1434
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1437
def get_options(self, version_id):
1438
"""Return a string represention options.
1442
return self._cache[version_id][1]
1444
def get_parents(self, version_id):
1445
"""Return parents of specified version ignoring ghosts."""
1446
return [parent for parent in self._cache[version_id][4]
1447
if parent in self._cache]
1449
def get_parents_with_ghosts(self, version_id):
1450
"""Return parents of specified version with ghosts."""
1451
return self._cache[version_id][4]
1453
def check_versions_present(self, version_ids):
1454
"""Check that all specified versions are present."""
1456
for version_id in version_ids:
1457
if version_id not in cache:
1458
raise RevisionNotPresent(version_id, self._filename)
1461
class KnitGraphIndex(object):
1462
"""A knit index that builds on GraphIndex."""
1464
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1465
"""Construct a KnitGraphIndex on a graph_index.
1467
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1468
:param deltas: Allow delta-compressed records.
1469
:param add_callback: If not None, allow additions to the index and call
1470
this callback with a list of added GraphIndex nodes:
1471
[(node, value, node_refs), ...]
1472
:param parents: If True, record knits parents, if not do not record
1475
self._graph_index = graph_index
1476
self._deltas = deltas
1477
self._add_callback = add_callback
1478
self._parents = parents
1479
if deltas and not parents:
1480
raise KnitCorrupt(self, "Cannot do delta compression without "
1483
def _get_entries(self, keys, check_present=False):
1484
"""Get the entries for keys.
1486
:param keys: An iterable of index keys, - 1-tuples.
1491
for node in self._graph_index.iter_entries(keys):
1493
found_keys.add(node[1])
1495
# adapt parentless index to the rest of the code.
1496
for node in self._graph_index.iter_entries(keys):
1497
yield node[0], node[1], node[2], ()
1498
found_keys.add(node[1])
1500
missing_keys = keys.difference(found_keys)
1502
raise RevisionNotPresent(missing_keys.pop(), self)
1504
def _present_keys(self, version_ids):
1506
node[1] for node in self._get_entries(version_ids)])
1508
def _parentless_ancestry(self, versions):
1509
"""Honour the get_ancestry API for parentless knit indices."""
1510
wanted_keys = self._version_ids_to_keys(versions)
1511
present_keys = self._present_keys(wanted_keys)
1512
missing = set(wanted_keys).difference(present_keys)
1514
raise RevisionNotPresent(missing.pop(), self)
1515
return list(self._keys_to_version_ids(present_keys))
1517
def get_ancestry(self, versions, topo_sorted=True):
1518
"""See VersionedFile.get_ancestry."""
1519
if not self._parents:
1520
return self._parentless_ancestry(versions)
1521
# XXX: This will do len(history) index calls - perhaps
1522
# it should be altered to be a index core feature?
1523
# get a graph of all the mentioned versions:
1526
versions = self._version_ids_to_keys(versions)
1527
pending = set(versions)
1529
# get all pending nodes
1530
this_iteration = pending
1531
new_nodes = self._get_entries(this_iteration)
1534
for (index, key, value, node_refs) in new_nodes:
1535
# dont ask for ghosties - otherwise
1536
# we we can end up looping with pending
1537
# being entirely ghosted.
1538
graph[key] = [parent for parent in node_refs[0]
1539
if parent not in ghosts]
1541
for parent in graph[key]:
1542
# dont examine known nodes again
1547
ghosts.update(this_iteration.difference(found))
1548
if versions.difference(graph):
1549
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1551
result_keys = topo_sort(graph.items())
1553
result_keys = graph.iterkeys()
1554
return [key[0] for key in result_keys]
1556
def get_ancestry_with_ghosts(self, versions):
1557
"""See VersionedFile.get_ancestry."""
1558
if not self._parents:
1559
return self._parentless_ancestry(versions)
1560
# XXX: This will do len(history) index calls - perhaps
1561
# it should be altered to be a index core feature?
1562
# get a graph of all the mentioned versions:
1564
versions = self._version_ids_to_keys(versions)
1565
pending = set(versions)
1567
# get all pending nodes
1568
this_iteration = pending
1569
new_nodes = self._get_entries(this_iteration)
1571
for (index, key, value, node_refs) in new_nodes:
1572
graph[key] = node_refs[0]
1574
for parent in graph[key]:
1575
# dont examine known nodes again
1579
missing_versions = this_iteration.difference(graph)
1580
missing_needed = versions.intersection(missing_versions)
1582
raise RevisionNotPresent(missing_needed.pop(), self)
1583
for missing_version in missing_versions:
1584
# add a key, no parents
1585
graph[missing_version] = []
1586
pending.discard(missing_version) # don't look for it
1587
result_keys = topo_sort(graph.items())
1588
return [key[0] for key in result_keys]
1590
def get_graph(self):
1591
"""Return a list of the node:parents lists from this knit index."""
1592
if not self._parents:
1593
return [(key, ()) for key in self.get_versions()]
1595
for index, key, value, refs in self._graph_index.iter_all_entries():
1596
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1599
def iter_parents(self, version_ids):
1600
"""Iterate through the parents for many version ids.
1602
:param version_ids: An iterable yielding version_ids.
1603
:return: An iterator that yields (version_id, parents). Requested
1604
version_ids not present in the versioned file are simply skipped.
1605
The order is undefined, allowing for different optimisations in
1606
the underlying implementation.
1609
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1611
present_parents = set()
1612
for node in all_nodes:
1613
all_parents.update(node[3][0])
1614
# any node we are querying must be present
1615
present_parents.add(node[1])
1616
unknown_parents = all_parents.difference(present_parents)
1617
present_parents.update(self._present_keys(unknown_parents))
1618
for node in all_nodes:
1620
for parent in node[3][0]:
1621
if parent in present_parents:
1622
parents.append(parent[0])
1623
yield node[1][0], tuple(parents)
1625
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1626
yield node[1][0], ()
1628
def num_versions(self):
1629
return len(list(self._graph_index.iter_all_entries()))
1631
__len__ = num_versions
1633
def get_versions(self):
1634
"""Get all the versions in the file. not topologically sorted."""
1635
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1637
def has_version(self, version_id):
1638
"""True if the version is in the index."""
1639
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1641
def _keys_to_version_ids(self, keys):
1642
return tuple(key[0] for key in keys)
1644
def get_position(self, version_id):
1645
"""Return details needed to access the version.
1647
:return: a tuple (index, data position, size) to hand to the access
1648
logic to get the record.
1650
node = self._get_node(version_id)
1651
bits = node[2][1:].split(' ')
1652
return node[0], int(bits[0]), int(bits[1])
1654
def get_method(self, version_id):
1655
"""Return compression method of specified version."""
1656
if not self._deltas:
1658
return self._parent_compression(self._get_node(version_id)[3][1])
1660
def _parent_compression(self, reference_list):
1661
# use the second reference list to decide if this is delta'd or not.
1662
if len(reference_list):
1667
def _get_node(self, version_id):
1668
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1670
def get_options(self, version_id):
1671
"""Return a string represention options.
1675
node = self._get_node(version_id)
1676
if not self._deltas:
1677
options = ['fulltext']
1679
options = [self._parent_compression(node[3][1])]
1680
if node[2][0] == 'N':
1681
options.append('no-eol')
1684
def get_parents(self, version_id):
1685
"""Return parents of specified version ignoring ghosts."""
1686
parents = list(self.iter_parents([version_id]))
1689
raise errors.RevisionNotPresent(version_id, self)
1690
return parents[0][1]
1692
def get_parents_with_ghosts(self, version_id):
1693
"""Return parents of specified version with ghosts."""
1694
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1695
check_present=True))
1696
if not self._parents:
1698
return self._keys_to_version_ids(nodes[0][3][0])
1700
def check_versions_present(self, version_ids):
1701
"""Check that all specified versions are present."""
1702
keys = self._version_ids_to_keys(version_ids)
1703
present = self._present_keys(keys)
1704
missing = keys.difference(present)
1706
raise RevisionNotPresent(missing.pop(), self)
1708
def add_version(self, version_id, options, access_memo, parents):
1709
"""Add a version record to the index."""
1710
return self.add_versions(((version_id, options, access_memo, parents),))
1712
def add_versions(self, versions):
1713
"""Add multiple versions to the index.
1715
This function does not insert data into the Immutable GraphIndex
1716
backing the KnitGraphIndex, instead it prepares data for insertion by
1717
the caller and checks that it is safe to insert then calls
1718
self._add_callback with the prepared GraphIndex nodes.
1720
:param versions: a list of tuples:
1721
(version_id, options, pos, size, parents).
1723
if not self._add_callback:
1724
raise errors.ReadOnlyError(self)
1725
# we hope there are no repositories with inconsistent parentage
1730
for (version_id, options, access_memo, parents) in versions:
1731
index, pos, size = access_memo
1732
key = (version_id, )
1733
parents = tuple((parent, ) for parent in parents)
1734
if 'no-eol' in options:
1738
value += "%d %d" % (pos, size)
1739
if not self._deltas:
1740
if 'line-delta' in options:
1741
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1744
if 'line-delta' in options:
1745
node_refs = (parents, (parents[0],))
1747
node_refs = (parents, ())
1749
node_refs = (parents, )
1752
raise KnitCorrupt(self, "attempt to add node with parents "
1753
"in parentless index.")
1755
keys[key] = (value, node_refs)
1756
present_nodes = self._get_entries(keys)
1757
for (index, key, value, node_refs) in present_nodes:
1758
if (value, node_refs) != keys[key]:
1759
raise KnitCorrupt(self, "inconsistent details in add_versions"
1760
": %s %s" % ((value, node_refs), keys[key]))
1764
for key, (value, node_refs) in keys.iteritems():
1765
result.append((key, value, node_refs))
1767
for key, (value, node_refs) in keys.iteritems():
1768
result.append((key, value))
1769
self._add_callback(result)
1771
def _version_ids_to_keys(self, version_ids):
1772
return set((version_id, ) for version_id in version_ids)
1775
class _KnitAccess(object):
1776
"""Access to knit records in a .knit file."""
1778
def __init__(self, transport, filename, _file_mode, _dir_mode,
1779
_need_to_create, _create_parent_dir):
1780
"""Create a _KnitAccess for accessing and inserting data.
1782
:param transport: The transport the .knit is located on.
1783
:param filename: The filename of the .knit.
1785
self._transport = transport
1786
self._filename = filename
1787
self._file_mode = _file_mode
1788
self._dir_mode = _dir_mode
1789
self._need_to_create = _need_to_create
1790
self._create_parent_dir = _create_parent_dir
1792
def add_raw_records(self, sizes, raw_data):
1793
"""Add raw knit bytes to a storage area.
1795
The data is spooled to whereever the access method is storing data.
1797
:param sizes: An iterable containing the size of each raw data segment.
1798
:param raw_data: A bytestring containing the data.
1799
:return: A list of memos to retrieve the record later. Each memo is a
1800
tuple - (index, pos, length), where the index field is always None
1801
for the .knit access method.
1803
assert type(raw_data) == str, \
1804
'data must be plain bytes was %s' % type(raw_data)
1805
if not self._need_to_create:
1806
base = self._transport.append_bytes(self._filename, raw_data)
1808
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1809
create_parent_dir=self._create_parent_dir,
1810
mode=self._file_mode,
1811
dir_mode=self._dir_mode)
1812
self._need_to_create = False
1816
result.append((None, base, size))
1821
"""IFF this data access has its own storage area, initialise it.
1825
self._transport.put_bytes_non_atomic(self._filename, '',
1826
mode=self._file_mode)
1828
def open_file(self):
1829
"""IFF this data access can be represented as a single file, open it.
1831
For knits that are not mapped to a single file on disk this will
1834
:return: None or a file handle.
1837
return self._transport.get(self._filename)
1842
def get_raw_records(self, memos_for_retrieval):
1843
"""Get the raw bytes for a records.
1845
:param memos_for_retrieval: An iterable containing the (index, pos,
1846
length) memo for retrieving the bytes. The .knit method ignores
1847
the index as there is always only a single file.
1848
:return: An iterator over the bytes of the records.
1850
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1851
for pos, data in self._transport.readv(self._filename, read_vector):
1855
class _PackAccess(object):
1856
"""Access to knit records via a collection of packs."""
1858
def __init__(self, index_to_packs, writer=None):
1859
"""Create a _PackAccess object.
1861
:param index_to_packs: A dict mapping index objects to the transport
1862
and file names for obtaining data.
1863
:param writer: A tuple (pack.ContainerWriter, write_index) which
1864
contains the pack to write, and the index that reads from it will
1868
self.container_writer = writer[0]
1869
self.write_index = writer[1]
1871
self.container_writer = None
1872
self.write_index = None
1873
self.indices = index_to_packs
1875
def add_raw_records(self, sizes, raw_data):
1876
"""Add raw knit bytes to a storage area.
1878
The data is spooled to the container writer in one bytes-record per
1881
:param sizes: An iterable containing the size of each raw data segment.
1882
:param raw_data: A bytestring containing the data.
1883
:return: A list of memos to retrieve the record later. Each memo is a
1884
tuple - (index, pos, length), where the index field is the
1885
write_index object supplied to the PackAccess object.
1887
assert type(raw_data) == str, \
1888
'data must be plain bytes was %s' % type(raw_data)
1892
p_offset, p_length = self.container_writer.add_bytes_record(
1893
raw_data[offset:offset+size], [])
1895
result.append((self.write_index, p_offset, p_length))
1899
"""Pack based knits do not get individually created."""
1901
def get_raw_records(self, memos_for_retrieval):
1902
"""Get the raw bytes for a records.
1904
:param memos_for_retrieval: An iterable containing the (index, pos,
1905
length) memo for retrieving the bytes. The Pack access method
1906
looks up the pack to use for a given record in its index_to_pack
1908
:return: An iterator over the bytes of the records.
1910
# first pass, group into same-index requests
1912
current_index = None
1913
for (index, offset, length) in memos_for_retrieval:
1914
if current_index == index:
1915
current_list.append((offset, length))
1917
if current_index is not None:
1918
request_lists.append((current_index, current_list))
1919
current_index = index
1920
current_list = [(offset, length)]
1921
# handle the last entry
1922
if current_index is not None:
1923
request_lists.append((current_index, current_list))
1924
for index, offsets in request_lists:
1925
transport, path = self.indices[index]
1926
reader = pack.make_readv_reader(transport, path, offsets)
1927
for names, read_func in reader.iter_records():
1928
yield read_func(None)
1930
def open_file(self):
1931
"""Pack based knits have no single file."""
1934
def set_writer(self, writer, index, (transport, packname)):
1935
"""Set a writer to use for adding data."""
1936
self.indices[index] = (transport, packname)
1937
self.container_writer = writer
1938
self.write_index = index
1941
class _KnitData(object):
1942
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1944
The KnitData class provides the logic for parsing and using knit records,
1945
making use of an access method for the low level read and write operations.
1948
def __init__(self, access):
1949
"""Create a KnitData object.
1951
:param access: The access method to use. Access methods such as
1952
_KnitAccess manage the insertion of raw records and the subsequent
1953
retrieval of the same.
1955
self._access = access
1956
self._checked = False
1957
# TODO: jam 20060713 conceptually, this could spill to disk
1958
# if the cached size gets larger than a certain amount
1959
# but it complicates the model a bit, so for now just use
1960
# a simple dictionary
1962
self._do_cache = False
1964
def enable_cache(self):
1965
"""Enable caching of reads."""
1966
self._do_cache = True
1968
def clear_cache(self):
1969
"""Clear the record cache."""
1970
self._do_cache = False
1973
def _open_file(self):
1974
return self._access.open_file()
1976
def _record_to_data(self, version_id, digest, lines):
1977
"""Convert version_id, digest, lines into a raw data block.
1979
:return: (len, a StringIO instance with the raw data ready to read.)
1982
data_file = GzipFile(None, mode='wb', fileobj=sio,
1983
compresslevel=Z_DEFAULT_COMPRESSION)
1985
assert isinstance(version_id, str)
1986
data_file.writelines(chain(
1987
["version %s %d %s\n" % (version_id,
1991
["end %s\n" % version_id]))
1998
def add_raw_records(self, sizes, raw_data):
1999
"""Append a prepared record to the data file.
2001
:param sizes: An iterable containing the size of each raw data segment.
2002
:param raw_data: A bytestring containing the data.
2003
:return: a list of index data for the way the data was stored.
2004
See the access method add_raw_records documentation for more
2007
return self._access.add_raw_records(sizes, raw_data)
2009
def add_record(self, version_id, digest, lines):
2010
"""Write new text record to disk.
2012
Returns index data for retrieving it later, as per add_raw_records.
2014
size, sio = self._record_to_data(version_id, digest, lines)
2015
result = self.add_raw_records([size], sio.getvalue())
2017
self._cache[version_id] = sio.getvalue()
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
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2191
pb = ui.ui_factory.nested_progress_bar()
2193
version_ids = list(version_ids)
2194
if None in version_ids:
2195
version_ids.remove(None)
2197
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2198
this_versions = set(self.target._index.get_versions())
2199
# XXX: For efficiency we should not look at the whole index,
2200
# we only need to consider the referenced revisions - they
2201
# must all be present, or the method must be full-text.
2202
# TODO, RBC 20070919
2203
needed_versions = self.source_ancestry - this_versions
2205
if not needed_versions:
2207
full_list = topo_sort(self.source.get_graph())
2209
version_list = [i for i in full_list if (not self.target.has_version(i)
2210
and i in needed_versions)]
2214
copy_queue_records = []
2216
for version_id in version_list:
2217
options = self.source._index.get_options(version_id)
2218
parents = self.source._index.get_parents_with_ghosts(version_id)
2219
# check that its will be a consistent copy:
2220
for parent in parents:
2221
# if source has the parent, we must :
2222
# * already have it or
2223
# * have it scheduled already
2224
# otherwise we don't care
2225
assert (self.target.has_version(parent) or
2226
parent in copy_set or
2227
not self.source.has_version(parent))
2228
index_memo = self.source._index.get_position(version_id)
2229
copy_queue_records.append((version_id, index_memo))
2230
copy_queue.append((version_id, options, parents))
2231
copy_set.add(version_id)
2233
# data suck the join:
2235
total = len(version_list)
2238
for (version_id, raw_data), \
2239
(version_id2, options, parents) in \
2240
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2242
assert version_id == version_id2, 'logic error, inconsistent results'
2244
pb.update("Joining knit", count, total)
2245
raw_records.append((version_id, options, parents, len(raw_data)))
2246
raw_datum.append(raw_data)
2247
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2253
InterVersionedFile.register_optimiser(InterKnit)
2256
class WeaveToKnit(InterVersionedFile):
2257
"""Optimised code paths for weave to knit operations."""
2259
_matching_file_from_factory = bzrlib.weave.WeaveFile
2260
_matching_file_to_factory = KnitVersionedFile
2263
def is_compatible(source, target):
2264
"""Be compatible with weaves to knits."""
2266
return (isinstance(source, bzrlib.weave.Weave) and
2267
isinstance(target, KnitVersionedFile))
2268
except AttributeError:
2271
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2272
"""See InterVersionedFile.join."""
2273
assert isinstance(self.source, bzrlib.weave.Weave)
2274
assert isinstance(self.target, KnitVersionedFile)
2276
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2281
pb = ui.ui_factory.nested_progress_bar()
2283
version_ids = list(version_ids)
2285
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2286
this_versions = set(self.target._index.get_versions())
2287
needed_versions = self.source_ancestry - this_versions
2289
if not needed_versions:
2291
full_list = topo_sort(self.source.get_graph())
2293
version_list = [i for i in full_list if (not self.target.has_version(i)
2294
and i in needed_versions)]
2298
total = len(version_list)
2299
for version_id in version_list:
2300
pb.update("Converting to knit", count, total)
2301
parents = self.source.get_parents(version_id)
2302
# check that its will be a consistent copy:
2303
for parent in parents:
2304
# if source has the parent, we must already have it
2305
assert (self.target.has_version(parent))
2306
self.target.add_lines(
2307
version_id, parents, self.source.get_lines(version_id))
2314
InterVersionedFile.register_optimiser(WeaveToKnit)
2317
# Deprecated, use PatienceSequenceMatcher instead
2318
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2321
def annotate_knit(knit, revision_id):
2322
"""Annotate a knit with no cached annotations.
2324
This implementation is for knits with no cached annotations.
2325
It will work for knits with cached annotations, but this is not
2328
ancestry = knit.get_ancestry(revision_id)
2329
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2331
for candidate in ancestry:
2332
if candidate in annotations:
2334
parents = knit.get_parents(candidate)
2335
if len(parents) == 0:
2337
elif knit._index.get_method(candidate) != 'line-delta':
2340
parent, sha1, noeol, delta = knit.get_delta(candidate)
2341
blocks = KnitContent.get_line_delta_blocks(delta,
2342
fulltext[parents[0]], fulltext[candidate])
2343
annotations[candidate] = list(annotate.reannotate([annotations[p]
2344
for p in parents], fulltext[candidate], candidate, blocks))
2345
return iter(annotations[revision_id])
2349
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2351
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
