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)
198
return [text for origin, text in self._lines]
199
except ValueError, e:
200
# most commonly (only?) caused by the internal form of the knit
201
# missing annotation information because of a bug - see thread
203
raise KnitCorrupt(self,
204
"line in annotated knit missing annotation information: %s"
208
return AnnotatedKnitContent(self._lines[:])
211
class PlainKnitContent(KnitContent):
212
"""Unannotated content.
214
When annotate[_iter] is called on this content, the same version is reported
215
for all lines. Generally, annotate[_iter] is not useful on PlainKnitContent
219
def __init__(self, lines, version_id):
221
self._version_id = version_id
223
def annotate_iter(self):
224
"""Yield tuples of (origin, text) for each content line."""
225
for line in self._lines:
226
yield self._version_id, line
229
return PlainKnitContent(self._lines[:], self._version_id)
231
def strip_last_line_newline(self):
232
self._lines[-1] = self._lines[-1].rstrip('\n')
238
class KnitAnnotateFactory(object):
239
"""Factory for creating annotated Content objects."""
243
def make(self, lines, version_id):
244
num_lines = len(lines)
245
return AnnotatedKnitContent(zip([version_id] * num_lines, lines))
247
def parse_fulltext(self, content, version_id):
248
"""Convert fulltext to internal representation
250
fulltext content is of the format
251
revid(utf8) plaintext\n
252
internal representation is of the format:
255
# TODO: jam 20070209 The tests expect this to be returned as tuples,
256
# but the code itself doesn't really depend on that.
257
# Figure out a way to not require the overhead of turning the
258
# list back into tuples.
259
lines = [tuple(line.split(' ', 1)) for line in content]
260
return AnnotatedKnitContent(lines)
262
def parse_line_delta_iter(self, lines):
263
return iter(self.parse_line_delta(lines))
265
def parse_line_delta(self, lines, version_id, plain=False):
266
"""Convert a line based delta into internal representation.
268
line delta is in the form of:
269
intstart intend intcount
271
revid(utf8) newline\n
272
internal representation is
273
(start, end, count, [1..count tuples (revid, newline)])
275
:param plain: If True, the lines are returned as a plain
276
list without annotations, not as a list of (origin, content) tuples, i.e.
277
(start, end, count, [1..count newline])
284
def cache_and_return(line):
285
origin, text = line.split(' ', 1)
286
return cache.setdefault(origin, origin), text
288
# walk through the lines parsing.
289
# Note that the plain test is explicitly pulled out of the
290
# loop to minimise any performance impact
293
start, end, count = [int(n) for n in header.split(',')]
294
contents = [next().split(' ', 1)[1] for i in xrange(count)]
295
result.append((start, end, count, contents))
298
start, end, count = [int(n) for n in header.split(',')]
299
contents = [tuple(next().split(' ', 1)) for i in xrange(count)]
300
result.append((start, end, count, contents))
303
def get_fulltext_content(self, lines):
304
"""Extract just the content lines from a fulltext."""
305
return (line.split(' ', 1)[1] for line in lines)
307
def get_linedelta_content(self, lines):
308
"""Extract just the content from a line delta.
310
This doesn't return all of the extra information stored in a delta.
311
Only the actual content lines.
316
header = header.split(',')
317
count = int(header[2])
318
for i in xrange(count):
319
origin, text = next().split(' ', 1)
322
def lower_fulltext(self, content):
323
"""convert a fulltext content record into a serializable form.
325
see parse_fulltext which this inverts.
327
# TODO: jam 20070209 We only do the caching thing to make sure that
328
# the origin is a valid utf-8 line, eventually we could remove it
329
return ['%s %s' % (o, t) for o, t in content._lines]
331
def lower_line_delta(self, delta):
332
"""convert a delta into a serializable form.
334
See parse_line_delta which this inverts.
336
# TODO: jam 20070209 We only do the caching thing to make sure that
337
# the origin is a valid utf-8 line, eventually we could remove it
339
for start, end, c, lines in delta:
340
out.append('%d,%d,%d\n' % (start, end, c))
341
out.extend(origin + ' ' + text
342
for origin, text in lines)
345
def annotate_iter(self, knit, version_id):
346
content = knit._get_content(version_id)
347
return content.annotate_iter()
350
class KnitPlainFactory(object):
351
"""Factory for creating plain Content objects."""
355
def make(self, lines, version_id):
356
return PlainKnitContent(lines, version_id)
358
def parse_fulltext(self, content, version_id):
359
"""This parses an unannotated fulltext.
361
Note that this is not a noop - the internal representation
362
has (versionid, line) - its just a constant versionid.
364
return self.make(content, version_id)
366
def parse_line_delta_iter(self, lines, version_id):
368
num_lines = len(lines)
369
while cur < num_lines:
372
start, end, c = [int(n) for n in header.split(',')]
373
yield start, end, c, lines[cur:cur+c]
376
def parse_line_delta(self, lines, version_id):
377
return list(self.parse_line_delta_iter(lines, version_id))
379
def get_fulltext_content(self, lines):
380
"""Extract just the content lines from a fulltext."""
383
def get_linedelta_content(self, lines):
384
"""Extract just the content from a line delta.
386
This doesn't return all of the extra information stored in a delta.
387
Only the actual content lines.
392
header = header.split(',')
393
count = int(header[2])
394
for i in xrange(count):
397
def lower_fulltext(self, content):
398
return content.text()
400
def lower_line_delta(self, delta):
402
for start, end, c, lines in delta:
403
out.append('%d,%d,%d\n' % (start, end, c))
407
def annotate_iter(self, knit, version_id):
408
return annotate_knit(knit, version_id)
411
def make_empty_knit(transport, relpath):
412
"""Construct a empty knit at the specified location."""
413
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
416
class KnitVersionedFile(VersionedFile):
417
"""Weave-like structure with faster random access.
419
A knit stores a number of texts and a summary of the relationships
420
between them. Texts are identified by a string version-id. Texts
421
are normally stored and retrieved as a series of lines, but can
422
also be passed as single strings.
424
Lines are stored with the trailing newline (if any) included, to
425
avoid special cases for files with no final newline. Lines are
426
composed of 8-bit characters, not unicode. The combination of
427
these approaches should mean any 'binary' file can be safely
428
stored and retrieved.
431
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
432
factory=None, delta=True, create=False, create_parent_dir=False,
433
delay_create=False, dir_mode=None, index=None, access_method=None):
434
"""Construct a knit at location specified by relpath.
436
:param create: If not True, only open an existing knit.
437
:param create_parent_dir: If True, create the parent directory if
438
creating the file fails. (This is used for stores with
439
hash-prefixes that may not exist yet)
440
:param delay_create: The calling code is aware that the knit won't
441
actually be created until the first data is stored.
442
:param index: An index to use for the knit.
444
if access_mode is None:
446
super(KnitVersionedFile, self).__init__(access_mode)
447
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
448
self.transport = transport
449
self.filename = relpath
450
self.factory = factory or KnitAnnotateFactory()
451
self.writable = (access_mode == 'w')
454
self._max_delta_chain = 200
457
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
458
access_mode, create=create, file_mode=file_mode,
459
create_parent_dir=create_parent_dir, delay_create=delay_create,
463
if access_method is None:
464
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
465
((create and not len(self)) and delay_create), create_parent_dir)
467
_access = access_method
468
if create and not len(self) and not delay_create:
470
self._data = _KnitData(_access)
473
return '%s(%s)' % (self.__class__.__name__,
474
self.transport.abspath(self.filename))
476
def _check_should_delta(self, first_parents):
477
"""Iterate back through the parent listing, looking for a fulltext.
479
This is used when we want to decide whether to add a delta or a new
480
fulltext. It searches for _max_delta_chain parents. When it finds a
481
fulltext parent, it sees if the total size of the deltas leading up to
482
it is large enough to indicate that we want a new full text anyway.
484
Return True if we should create a new delta, False if we should use a
489
delta_parents = first_parents
490
for count in xrange(self._max_delta_chain):
491
parent = delta_parents[0]
492
method = self._index.get_method(parent)
493
index, pos, size = self._index.get_position(parent)
494
if method == 'fulltext':
498
delta_parents = self._index.get_parents(parent)
500
# We couldn't find a fulltext, so we must create a new one
503
return fulltext_size > delta_size
505
def _add_raw_records(self, records, data):
506
"""Add all the records 'records' with data pre-joined in 'data'.
508
:param records: A list of tuples(version_id, options, parents, size).
509
:param data: The data for the records. When it is written, the records
510
are adjusted to have pos pointing into data by the sum of
511
the preceding records sizes.
514
raw_record_sizes = [record[3] for record in records]
515
positions = self._data.add_raw_records(raw_record_sizes, data)
518
for (version_id, options, parents, size), access_memo in zip(
520
index_entries.append((version_id, options, access_memo, parents))
521
if self._data._do_cache:
522
self._data._cache[version_id] = data[offset:offset+size]
524
self._index.add_versions(index_entries)
526
def enable_cache(self):
527
"""Start caching data for this knit"""
528
self._data.enable_cache()
530
def clear_cache(self):
531
"""Clear the data cache only."""
532
self._data.clear_cache()
534
def copy_to(self, name, transport):
535
"""See VersionedFile.copy_to()."""
536
# copy the current index to a temp index to avoid racing with local
538
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
539
self.transport.get(self._index._filename))
541
f = self._data._open_file()
543
transport.put_file(name + DATA_SUFFIX, f)
546
# move the copied index into place
547
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
549
def create_empty(self, name, transport, mode=None):
550
return KnitVersionedFile(name, transport, factory=self.factory,
551
delta=self.delta, create=True)
553
def get_data_stream(self, required_versions):
554
"""Get a data stream for the specified versions.
556
Versions may be returned in any order, not necessarily the order
559
:param required_versions: The exact set of versions to be extracted.
560
Unlike some other knit methods, this is not used to generate a
561
transitive closure, rather it is used precisely as given.
563
:returns: format_signature, list of (version, options, length, parents),
566
if not isinstance(required_versions, set):
567
required_versions = set(required_versions)
568
# we don't care about inclusions, the caller cares.
569
# but we need to setup a list of records to visit.
570
for version_id in required_versions:
571
if not self.has_version(version_id):
572
raise RevisionNotPresent(version_id, self.filename)
573
# Pick the desired versions out of the index in oldest-to-newest order
575
for version_id in self.versions():
576
if version_id in required_versions:
577
version_list.append(version_id)
579
# create the list of version information for the result
580
copy_queue_records = []
582
result_version_list = []
583
for version_id in version_list:
584
options = self._index.get_options(version_id)
585
parents = self._index.get_parents_with_ghosts(version_id)
586
index_memo = self._index.get_position(version_id)
587
copy_queue_records.append((version_id, index_memo))
588
none, data_pos, data_size = index_memo
589
copy_set.add(version_id)
590
# version, options, length, parents
591
result_version_list.append((version_id, options, data_size,
594
# Read the compressed record data.
596
# From here down to the return should really be logic in the returned
597
# callable -- in a class that adapts read_records_iter_raw to read
600
for (version_id, raw_data), \
601
(version_id2, options, _, parents) in \
602
izip(self._data.read_records_iter_raw(copy_queue_records),
603
result_version_list):
604
assert version_id == version_id2, 'logic error, inconsistent results'
605
raw_datum.append(raw_data)
606
pseudo_file = StringIO(''.join(raw_datum))
609
return pseudo_file.read()
611
return pseudo_file.read(length)
612
return (self.get_format_signature(), result_version_list, read)
614
def _extract_blocks(self, version_id, source, target):
615
if self._index.get_method(version_id) != 'line-delta':
617
parent, sha1, noeol, delta = self.get_delta(version_id)
618
return KnitContent.get_line_delta_blocks(delta, source, target)
620
def get_delta(self, version_id):
621
"""Get a delta for constructing version from some other version."""
622
self.check_not_reserved_id(version_id)
623
parents = self.get_parents(version_id)
628
index_memo = self._index.get_position(version_id)
629
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
630
noeol = 'no-eol' in self._index.get_options(version_id)
631
if 'fulltext' == self._index.get_method(version_id):
632
new_content = self.factory.parse_fulltext(data, version_id)
633
if parent is not None:
634
reference_content = self._get_content(parent)
635
old_texts = reference_content.text()
638
new_texts = new_content.text()
639
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
641
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
643
delta = self.factory.parse_line_delta(data, version_id)
644
return parent, sha1, noeol, delta
646
def get_format_signature(self):
647
"""See VersionedFile.get_format_signature()."""
648
if self.factory.annotated:
649
annotated_part = "annotated"
651
annotated_part = "plain"
652
return "knit-%s" % (annotated_part,)
654
def get_graph_with_ghosts(self):
655
"""See VersionedFile.get_graph_with_ghosts()."""
656
graph_items = self._index.get_graph()
657
return dict(graph_items)
659
def get_sha1(self, version_id):
660
return self.get_sha1s([version_id])[0]
662
def get_sha1s(self, version_ids):
663
"""See VersionedFile.get_sha1()."""
664
record_map = self._get_record_map(version_ids)
665
# record entry 2 is the 'digest'.
666
return [record_map[v][2] for v in version_ids]
670
"""See VersionedFile.get_suffixes()."""
671
return [DATA_SUFFIX, INDEX_SUFFIX]
673
def has_ghost(self, version_id):
674
"""True if there is a ghost reference in the file to 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
if 'line-delta' in options:
727
# Make sure that this knit record is actually useful: a
728
# line-delta is no use unless we have its parent.
729
# Fetching from a broken repository with this problem
730
# shouldn't break the target repository.
731
if not self._index.has_version(parents[0]):
734
'line-delta from stream references '
735
'missing parent %s' % parents[0])
736
self._add_raw_records(
737
[(version_id, options, parents, length)],
738
reader_callable(length))
741
"""See VersionedFile.versions."""
742
if 'evil' in debug.debug_flags:
743
trace.mutter_callsite(2, "versions scales with size of history")
744
return self._index.get_versions()
746
def has_version(self, version_id):
747
"""See VersionedFile.has_version."""
748
if 'evil' in debug.debug_flags:
749
trace.mutter_callsite(2, "has_version is a LBYL scenario")
750
return self._index.has_version(version_id)
752
__contains__ = has_version
754
def _merge_annotations(self, content, parents, parent_texts={},
755
delta=None, annotated=None,
756
left_matching_blocks=None):
757
"""Merge annotations for content. This is done by comparing
758
the annotations based on changed to the text.
760
if left_matching_blocks is not None:
761
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
765
for parent_id in parents:
766
merge_content = self._get_content(parent_id, parent_texts)
767
if (parent_id == parents[0] and delta_seq is not None):
770
seq = patiencediff.PatienceSequenceMatcher(
771
None, merge_content.text(), content.text())
772
for i, j, n in seq.get_matching_blocks():
775
# this appears to copy (origin, text) pairs across to the
776
# new content for any line that matches the last-checked
778
content._lines[j:j+n] = merge_content._lines[i:i+n]
780
if delta_seq is None:
781
reference_content = self._get_content(parents[0], parent_texts)
782
new_texts = content.text()
783
old_texts = reference_content.text()
784
delta_seq = patiencediff.PatienceSequenceMatcher(
785
None, old_texts, new_texts)
786
return self._make_line_delta(delta_seq, content)
788
def _make_line_delta(self, delta_seq, new_content):
789
"""Generate a line delta from delta_seq and new_content."""
791
for op in delta_seq.get_opcodes():
794
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
797
def _get_components_positions(self, version_ids):
798
"""Produce a map of position data for the components of versions.
800
This data is intended to be used for retrieving the knit records.
802
A dict of version_id to (method, data_pos, data_size, next) is
804
method is the way referenced data should be applied.
805
data_pos is the position of the data in the knit.
806
data_size is the size of the data in the knit.
807
next is the build-parent of the version, or None for fulltexts.
810
for version_id in version_ids:
813
while cursor is not None and cursor not in component_data:
814
method = self._index.get_method(cursor)
815
if method == 'fulltext':
818
next = self.get_parents(cursor)[0]
819
index_memo = self._index.get_position(cursor)
820
component_data[cursor] = (method, index_memo, next)
822
return component_data
824
def _get_content(self, version_id, parent_texts={}):
825
"""Returns a content object that makes up the specified
827
cached_version = parent_texts.get(version_id, None)
828
if cached_version is not None:
829
if not self.has_version(version_id):
830
raise RevisionNotPresent(version_id, self.filename)
831
return cached_version
833
text_map, contents_map = self._get_content_maps([version_id])
834
return contents_map[version_id]
836
def _check_versions_present(self, version_ids):
837
"""Check that all specified versions are present."""
838
self._index.check_versions_present(version_ids)
840
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
841
nostore_sha, random_id, check_content):
842
"""See VersionedFile.add_lines_with_ghosts()."""
843
self._check_add(version_id, lines, random_id, check_content)
844
return self._add(version_id, lines, parents, self.delta,
845
parent_texts, None, nostore_sha, random_id)
847
def _add_lines(self, version_id, parents, lines, parent_texts,
848
left_matching_blocks, nostore_sha, random_id, check_content):
849
"""See VersionedFile.add_lines."""
850
self._check_add(version_id, lines, random_id, check_content)
851
self._check_versions_present(parents)
852
return self._add(version_id, lines[:], parents, self.delta,
853
parent_texts, left_matching_blocks, nostore_sha, random_id)
855
def _check_add(self, version_id, lines, random_id, check_content):
856
"""check that version_id and lines are safe to add."""
857
if contains_whitespace(version_id):
858
raise InvalidRevisionId(version_id, self.filename)
859
self.check_not_reserved_id(version_id)
860
# Technically this could be avoided if we are happy to allow duplicate
861
# id insertion when other things than bzr core insert texts, but it
862
# seems useful for folk using the knit api directly to have some safety
863
# blanket that we can disable.
864
if not random_id and self.has_version(version_id):
865
raise RevisionAlreadyPresent(version_id, self.filename)
867
self._check_lines_not_unicode(lines)
868
self._check_lines_are_lines(lines)
870
def _add(self, version_id, lines, parents, delta, parent_texts,
871
left_matching_blocks, nostore_sha, random_id):
872
"""Add a set of lines on top of version specified by parents.
874
If delta is true, compress the text as a line-delta against
877
Any versions not present will be converted into ghosts.
879
# first thing, if the content is something we don't need to store, find
881
line_bytes = ''.join(lines)
882
digest = sha_string(line_bytes)
883
if nostore_sha == digest:
884
raise errors.ExistingContent
887
if parent_texts is None:
889
for parent in parents:
890
if self.has_version(parent):
891
present_parents.append(parent)
893
# can only compress against the left most present parent.
895
(len(present_parents) == 0 or
896
present_parents[0] != parents[0])):
899
text_length = len(line_bytes)
902
if lines[-1][-1] != '\n':
903
# copy the contents of lines.
905
options.append('no-eol')
906
lines[-1] = lines[-1] + '\n'
910
# To speed the extract of texts the delta chain is limited
911
# to a fixed number of deltas. This should minimize both
912
# I/O and the time spend applying deltas.
913
delta = self._check_should_delta(present_parents)
915
assert isinstance(version_id, str)
916
content = self.factory.make(lines, version_id)
917
if delta or (self.factory.annotated and len(present_parents) > 0):
918
# Merge annotations from parent texts if needed.
919
delta_hunks = self._merge_annotations(content, present_parents,
920
parent_texts, delta, self.factory.annotated,
921
left_matching_blocks)
924
options.append('line-delta')
925
store_lines = self.factory.lower_line_delta(delta_hunks)
926
size, bytes = self._data._record_to_data(version_id, digest,
929
options.append('fulltext')
930
# isinstance is slower and we have no hierarchy.
931
if self.factory.__class__ == KnitPlainFactory:
932
# Use the already joined bytes saving iteration time in
934
size, bytes = self._data._record_to_data(version_id, digest,
937
# get mixed annotation + content and feed it into the
939
store_lines = self.factory.lower_fulltext(content)
940
size, bytes = self._data._record_to_data(version_id, digest,
943
access_memo = self._data.add_raw_records([size], bytes)[0]
944
self._index.add_versions(
945
((version_id, options, access_memo, parents),),
947
return digest, text_length, content
949
def check(self, progress_bar=None):
950
"""See VersionedFile.check()."""
952
def _clone_text(self, new_version_id, old_version_id, parents):
953
"""See VersionedFile.clone_text()."""
954
# FIXME RBC 20060228 make fast by only inserting an index with null
956
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
958
def get_lines(self, version_id):
959
"""See VersionedFile.get_lines()."""
960
return self.get_line_list([version_id])[0]
962
def _get_record_map(self, version_ids):
963
"""Produce a dictionary of knit records.
965
The keys are version_ids, the values are tuples of (method, content,
967
method is the way the content should be applied.
968
content is a KnitContent object.
969
digest is the SHA1 digest of this version id after all steps are done
970
next is the build-parent of the version, i.e. the leftmost ancestor.
971
If the method is fulltext, next will be None.
973
position_map = self._get_components_positions(version_ids)
974
# c = component_id, m = method, i_m = index_memo, n = next
975
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
977
for component_id, content, digest in \
978
self._data.read_records_iter(records):
979
method, index_memo, next = position_map[component_id]
980
record_map[component_id] = method, content, digest, next
984
def get_text(self, version_id):
985
"""See VersionedFile.get_text"""
986
return self.get_texts([version_id])[0]
988
def get_texts(self, version_ids):
989
return [''.join(l) for l in self.get_line_list(version_ids)]
991
def get_line_list(self, version_ids):
992
"""Return the texts of listed versions as a list of strings."""
993
for version_id in version_ids:
994
self.check_not_reserved_id(version_id)
995
text_map, content_map = self._get_content_maps(version_ids)
996
return [text_map[v] for v in version_ids]
998
_get_lf_split_line_list = get_line_list
1000
def _get_content_maps(self, version_ids):
1001
"""Produce maps of text and KnitContents
1003
:return: (text_map, content_map) where text_map contains the texts for
1004
the requested versions and content_map contains the KnitContents.
1005
Both dicts take version_ids as their keys.
1007
record_map = self._get_record_map(version_ids)
1012
for version_id in version_ids:
1015
while cursor is not None:
1016
method, data, digest, next = record_map[cursor]
1017
components.append((cursor, method, data, digest))
1018
if cursor in content_map:
1023
for component_id, method, data, digest in reversed(components):
1024
if component_id in content_map:
1025
content = content_map[component_id]
1027
if method == 'fulltext':
1028
assert content is None
1029
content = self.factory.parse_fulltext(data, version_id)
1030
elif method == 'line-delta':
1031
delta = self.factory.parse_line_delta(data, version_id)
1032
content = content.copy()
1033
content._lines = self._apply_delta(content._lines,
1035
content_map[component_id] = content
1037
if 'no-eol' in self._index.get_options(version_id):
1038
content = content.copy()
1039
content.strip_last_line_newline()
1040
final_content[version_id] = content
1042
# digest here is the digest from the last applied component.
1043
text = content.text()
1044
actual_sha = sha_strings(text)
1045
if actual_sha != digest:
1046
raise KnitCorrupt(self.filename,
1048
'\n of reconstructed text does not match'
1050
'\n for version %s' %
1051
(actual_sha, digest, version_id))
1052
text_map[version_id] = text
1053
return text_map, final_content
1056
def _apply_delta(lines, delta):
1057
"""Apply delta to lines."""
1060
for start, end, count, delta_lines in delta:
1061
lines[offset+start:offset+end] = delta_lines
1062
offset = offset + (start - end) + count
1065
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1067
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1068
if version_ids is None:
1069
version_ids = self.versions()
1071
pb = progress.DummyProgress()
1072
# we don't care about inclusions, the caller cares.
1073
# but we need to setup a list of records to visit.
1074
# we need version_id, position, length
1075
version_id_records = []
1076
requested_versions = set(version_ids)
1077
# filter for available versions
1078
for version_id in requested_versions:
1079
if not self.has_version(version_id):
1080
raise RevisionNotPresent(version_id, self.filename)
1081
# get a in-component-order queue:
1082
for version_id in self.versions():
1083
if version_id in requested_versions:
1084
index_memo = self._index.get_position(version_id)
1085
version_id_records.append((version_id, index_memo))
1087
total = len(version_id_records)
1088
for version_idx, (version_id, data, sha_value) in \
1089
enumerate(self._data.read_records_iter(version_id_records)):
1090
pb.update('Walking content.', version_idx, total)
1091
method = self._index.get_method(version_id)
1093
assert method in ('fulltext', 'line-delta')
1094
if method == 'fulltext':
1095
line_iterator = self.factory.get_fulltext_content(data)
1097
line_iterator = self.factory.get_linedelta_content(data)
1098
for line in line_iterator:
1101
pb.update('Walking content.', total, total)
1103
def iter_parents(self, version_ids):
1104
"""Iterate through the parents for many version ids.
1106
:param version_ids: An iterable yielding version_ids.
1107
:return: An iterator that yields (version_id, parents). Requested
1108
version_ids not present in the versioned file are simply skipped.
1109
The order is undefined, allowing for different optimisations in
1110
the underlying implementation.
1112
return self._index.iter_parents(version_ids)
1114
def num_versions(self):
1115
"""See VersionedFile.num_versions()."""
1116
return self._index.num_versions()
1118
__len__ = num_versions
1120
def annotate_iter(self, version_id):
1121
"""See VersionedFile.annotate_iter."""
1122
return self.factory.annotate_iter(self, version_id)
1124
def get_parents(self, version_id):
1125
"""See VersionedFile.get_parents."""
1128
# 52554 calls in 1264 872 internal down from 3674
1130
return self._index.get_parents(version_id)
1132
raise RevisionNotPresent(version_id, self.filename)
1134
def get_parents_with_ghosts(self, version_id):
1135
"""See VersionedFile.get_parents."""
1137
return self._index.get_parents_with_ghosts(version_id)
1139
raise RevisionNotPresent(version_id, self.filename)
1141
def get_ancestry(self, versions, topo_sorted=True):
1142
"""See VersionedFile.get_ancestry."""
1143
if isinstance(versions, basestring):
1144
versions = [versions]
1147
return self._index.get_ancestry(versions, topo_sorted)
1149
def get_ancestry_with_ghosts(self, versions):
1150
"""See VersionedFile.get_ancestry_with_ghosts."""
1151
if isinstance(versions, basestring):
1152
versions = [versions]
1155
return self._index.get_ancestry_with_ghosts(versions)
1157
def plan_merge(self, ver_a, ver_b):
1158
"""See VersionedFile.plan_merge."""
1159
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1160
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1161
annotated_a = self.annotate(ver_a)
1162
annotated_b = self.annotate(ver_b)
1163
return merge._plan_annotate_merge(annotated_a, annotated_b,
1164
ancestors_a, ancestors_b)
1167
class _KnitComponentFile(object):
1168
"""One of the files used to implement a knit database"""
1170
def __init__(self, transport, filename, mode, file_mode=None,
1171
create_parent_dir=False, dir_mode=None):
1172
self._transport = transport
1173
self._filename = filename
1175
self._file_mode = file_mode
1176
self._dir_mode = dir_mode
1177
self._create_parent_dir = create_parent_dir
1178
self._need_to_create = False
1180
def _full_path(self):
1181
"""Return the full path to this file."""
1182
return self._transport.base + self._filename
1184
def check_header(self, fp):
1185
line = fp.readline()
1187
# An empty file can actually be treated as though the file doesn't
1189
raise errors.NoSuchFile(self._full_path())
1190
if line != self.HEADER:
1191
raise KnitHeaderError(badline=line,
1192
filename=self._transport.abspath(self._filename))
1195
return '%s(%s)' % (self.__class__.__name__, self._filename)
1198
class _KnitIndex(_KnitComponentFile):
1199
"""Manages knit index file.
1201
The index is already kept in memory and read on startup, to enable
1202
fast lookups of revision information. The cursor of the index
1203
file is always pointing to the end, making it easy to append
1206
_cache is a cache for fast mapping from version id to a Index
1209
_history is a cache for fast mapping from indexes to version ids.
1211
The index data format is dictionary compressed when it comes to
1212
parent references; a index entry may only have parents that with a
1213
lover index number. As a result, the index is topological sorted.
1215
Duplicate entries may be written to the index for a single version id
1216
if this is done then the latter one completely replaces the former:
1217
this allows updates to correct version and parent information.
1218
Note that the two entries may share the delta, and that successive
1219
annotations and references MUST point to the first entry.
1221
The index file on disc contains a header, followed by one line per knit
1222
record. The same revision can be present in an index file more than once.
1223
The first occurrence gets assigned a sequence number starting from 0.
1225
The format of a single line is
1226
REVISION_ID FLAGS BYTE_OFFSET LENGTH( PARENT_ID|PARENT_SEQUENCE_ID)* :\n
1227
REVISION_ID is a utf8-encoded revision id
1228
FLAGS is a comma separated list of flags about the record. Values include
1229
no-eol, line-delta, fulltext.
1230
BYTE_OFFSET is the ascii representation of the byte offset in the data file
1231
that the the compressed data starts at.
1232
LENGTH is the ascii representation of the length of the data file.
1233
PARENT_ID a utf-8 revision id prefixed by a '.' that is a parent of
1235
PARENT_SEQUENCE_ID the ascii representation of the sequence number of a
1236
revision id already in the knit that is a parent of REVISION_ID.
1237
The ' :' marker is the end of record marker.
1240
when a write is interrupted to the index file, it will result in a line
1241
that does not end in ' :'. If the ' :' is not present at the end of a line,
1242
or at the end of the file, then the record that is missing it will be
1243
ignored by the parser.
1245
When writing new records to the index file, the data is preceded by '\n'
1246
to ensure that records always start on new lines even if the last write was
1247
interrupted. As a result its normal for the last line in the index to be
1248
missing a trailing newline. One can be added with no harmful effects.
1251
HEADER = "# bzr knit index 8\n"
1253
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1254
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1256
def _cache_version(self, version_id, options, pos, size, parents):
1257
"""Cache a version record in the history array and index cache.
1259
This is inlined into _load_data for performance. KEEP IN SYNC.
1260
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1263
# only want the _history index to reference the 1st index entry
1265
if version_id not in self._cache:
1266
index = len(self._history)
1267
self._history.append(version_id)
1269
index = self._cache[version_id][5]
1270
self._cache[version_id] = (version_id,
1277
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1278
create_parent_dir=False, delay_create=False, dir_mode=None):
1279
_KnitComponentFile.__init__(self, transport, filename, mode,
1280
file_mode=file_mode,
1281
create_parent_dir=create_parent_dir,
1284
# position in _history is the 'official' index for a revision
1285
# but the values may have come from a newer entry.
1286
# so - wc -l of a knit index is != the number of unique names
1290
fp = self._transport.get(self._filename)
1292
# _load_data may raise NoSuchFile if the target knit is
1294
_load_data(self, fp)
1298
if mode != 'w' or not create:
1301
self._need_to_create = True
1303
self._transport.put_bytes_non_atomic(
1304
self._filename, self.HEADER, mode=self._file_mode)
1306
def get_graph(self):
1307
"""Return a list of the node:parents lists from this knit index."""
1308
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1310
def get_ancestry(self, versions, topo_sorted=True):
1311
"""See VersionedFile.get_ancestry."""
1312
# get a graph of all the mentioned versions:
1314
pending = set(versions)
1317
version = pending.pop()
1320
parents = [p for p in cache[version][4] if p in cache]
1322
raise RevisionNotPresent(version, self._filename)
1323
# if not completed and not a ghost
1324
pending.update([p for p in parents if p not in graph])
1325
graph[version] = parents
1328
return topo_sort(graph.items())
1330
def get_ancestry_with_ghosts(self, versions):
1331
"""See VersionedFile.get_ancestry_with_ghosts."""
1332
# get a graph of all the mentioned versions:
1333
self.check_versions_present(versions)
1336
pending = set(versions)
1338
version = pending.pop()
1340
parents = cache[version][4]
1346
pending.update([p for p in parents if p not in graph])
1347
graph[version] = parents
1348
return topo_sort(graph.items())
1350
def iter_parents(self, version_ids):
1351
"""Iterate through the parents for many version ids.
1353
:param version_ids: An iterable yielding version_ids.
1354
:return: An iterator that yields (version_id, parents). Requested
1355
version_ids not present in the versioned file are simply skipped.
1356
The order is undefined, allowing for different optimisations in
1357
the underlying implementation.
1359
for version_id in version_ids:
1361
yield version_id, tuple(self.get_parents(version_id))
1365
def num_versions(self):
1366
return len(self._history)
1368
__len__ = num_versions
1370
def get_versions(self):
1371
"""Get all the versions in the file. not topologically sorted."""
1372
return self._history
1374
def _version_list_to_index(self, versions):
1377
for version in versions:
1378
if version in cache:
1379
# -- inlined lookup() --
1380
result_list.append(str(cache[version][5]))
1381
# -- end lookup () --
1383
result_list.append('.' + version)
1384
return ' '.join(result_list)
1386
def add_version(self, version_id, options, index_memo, parents):
1387
"""Add a version record to the index."""
1388
self.add_versions(((version_id, options, index_memo, parents),))
1390
def add_versions(self, versions, random_id=False):
1391
"""Add multiple versions to the index.
1393
:param versions: a list of tuples:
1394
(version_id, options, pos, size, parents).
1395
:param random_id: If True the ids being added were randomly generated
1396
and no check for existence will be performed.
1399
orig_history = self._history[:]
1400
orig_cache = self._cache.copy()
1403
for version_id, options, (index, pos, size), parents in versions:
1404
line = "\n%s %s %s %s %s :" % (version_id,
1408
self._version_list_to_index(parents))
1409
assert isinstance(line, str), \
1410
'content must be utf-8 encoded: %r' % (line,)
1412
self._cache_version(version_id, options, pos, size, parents)
1413
if not self._need_to_create:
1414
self._transport.append_bytes(self._filename, ''.join(lines))
1417
sio.write(self.HEADER)
1418
sio.writelines(lines)
1420
self._transport.put_file_non_atomic(self._filename, sio,
1421
create_parent_dir=self._create_parent_dir,
1422
mode=self._file_mode,
1423
dir_mode=self._dir_mode)
1424
self._need_to_create = False
1426
# If any problems happen, restore the original values and re-raise
1427
self._history = orig_history
1428
self._cache = orig_cache
1431
def has_version(self, version_id):
1432
"""True if the version is in the index."""
1433
return version_id in self._cache
1435
def get_position(self, version_id):
1436
"""Return details needed to access the version.
1438
.kndx indices do not support split-out data, so return None for the
1441
:return: a tuple (None, data position, size) to hand to the access
1442
logic to get the record.
1444
entry = self._cache[version_id]
1445
return None, entry[2], entry[3]
1447
def get_method(self, version_id):
1448
"""Return compression method of specified version."""
1450
options = self._cache[version_id][1]
1452
raise RevisionNotPresent(version_id, self._filename)
1453
if 'fulltext' in options:
1456
if 'line-delta' not in options:
1457
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1460
def get_options(self, version_id):
1461
"""Return a string represention options.
1465
return self._cache[version_id][1]
1467
def get_parents(self, version_id):
1468
"""Return parents of specified version ignoring ghosts."""
1469
return [parent for parent in self._cache[version_id][4]
1470
if parent in self._cache]
1472
def get_parents_with_ghosts(self, version_id):
1473
"""Return parents of specified version with ghosts."""
1474
return self._cache[version_id][4]
1476
def check_versions_present(self, version_ids):
1477
"""Check that all specified versions are present."""
1479
for version_id in version_ids:
1480
if version_id not in cache:
1481
raise RevisionNotPresent(version_id, self._filename)
1484
class KnitGraphIndex(object):
1485
"""A knit index that builds on GraphIndex."""
1487
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
1488
"""Construct a KnitGraphIndex on a graph_index.
1490
:param graph_index: An implementation of bzrlib.index.GraphIndex.
1491
:param deltas: Allow delta-compressed records.
1492
:param add_callback: If not None, allow additions to the index and call
1493
this callback with a list of added GraphIndex nodes:
1494
[(node, value, node_refs), ...]
1495
:param parents: If True, record knits parents, if not do not record
1498
self._graph_index = graph_index
1499
self._deltas = deltas
1500
self._add_callback = add_callback
1501
self._parents = parents
1502
if deltas and not parents:
1503
raise KnitCorrupt(self, "Cannot do delta compression without "
1506
def _get_entries(self, keys, check_present=False):
1507
"""Get the entries for keys.
1509
:param keys: An iterable of index keys, - 1-tuples.
1514
for node in self._graph_index.iter_entries(keys):
1516
found_keys.add(node[1])
1518
# adapt parentless index to the rest of the code.
1519
for node in self._graph_index.iter_entries(keys):
1520
yield node[0], node[1], node[2], ()
1521
found_keys.add(node[1])
1523
missing_keys = keys.difference(found_keys)
1525
raise RevisionNotPresent(missing_keys.pop(), self)
1527
def _present_keys(self, version_ids):
1529
node[1] for node in self._get_entries(version_ids)])
1531
def _parentless_ancestry(self, versions):
1532
"""Honour the get_ancestry API for parentless knit indices."""
1533
wanted_keys = self._version_ids_to_keys(versions)
1534
present_keys = self._present_keys(wanted_keys)
1535
missing = set(wanted_keys).difference(present_keys)
1537
raise RevisionNotPresent(missing.pop(), self)
1538
return list(self._keys_to_version_ids(present_keys))
1540
def get_ancestry(self, versions, topo_sorted=True):
1541
"""See VersionedFile.get_ancestry."""
1542
if not self._parents:
1543
return self._parentless_ancestry(versions)
1544
# XXX: This will do len(history) index calls - perhaps
1545
# it should be altered to be a index core feature?
1546
# get a graph of all the mentioned versions:
1549
versions = self._version_ids_to_keys(versions)
1550
pending = set(versions)
1552
# get all pending nodes
1553
this_iteration = pending
1554
new_nodes = self._get_entries(this_iteration)
1557
for (index, key, value, node_refs) in new_nodes:
1558
# dont ask for ghosties - otherwise
1559
# we we can end up looping with pending
1560
# being entirely ghosted.
1561
graph[key] = [parent for parent in node_refs[0]
1562
if parent not in ghosts]
1564
for parent in graph[key]:
1565
# dont examine known nodes again
1570
ghosts.update(this_iteration.difference(found))
1571
if versions.difference(graph):
1572
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1574
result_keys = topo_sort(graph.items())
1576
result_keys = graph.iterkeys()
1577
return [key[0] for key in result_keys]
1579
def get_ancestry_with_ghosts(self, versions):
1580
"""See VersionedFile.get_ancestry."""
1581
if not self._parents:
1582
return self._parentless_ancestry(versions)
1583
# XXX: This will do len(history) index calls - perhaps
1584
# it should be altered to be a index core feature?
1585
# get a graph of all the mentioned versions:
1587
versions = self._version_ids_to_keys(versions)
1588
pending = set(versions)
1590
# get all pending nodes
1591
this_iteration = pending
1592
new_nodes = self._get_entries(this_iteration)
1594
for (index, key, value, node_refs) in new_nodes:
1595
graph[key] = node_refs[0]
1597
for parent in graph[key]:
1598
# dont examine known nodes again
1602
missing_versions = this_iteration.difference(graph)
1603
missing_needed = versions.intersection(missing_versions)
1605
raise RevisionNotPresent(missing_needed.pop(), self)
1606
for missing_version in missing_versions:
1607
# add a key, no parents
1608
graph[missing_version] = []
1609
pending.discard(missing_version) # don't look for it
1610
result_keys = topo_sort(graph.items())
1611
return [key[0] for key in result_keys]
1613
def get_graph(self):
1614
"""Return a list of the node:parents lists from this knit index."""
1615
if not self._parents:
1616
return [(key, ()) for key in self.get_versions()]
1618
for index, key, value, refs in self._graph_index.iter_all_entries():
1619
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1622
def iter_parents(self, version_ids):
1623
"""Iterate through the parents for many version ids.
1625
:param version_ids: An iterable yielding version_ids.
1626
:return: An iterator that yields (version_id, parents). Requested
1627
version_ids not present in the versioned file are simply skipped.
1628
The order is undefined, allowing for different optimisations in
1629
the underlying implementation.
1632
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1634
present_parents = set()
1635
for node in all_nodes:
1636
all_parents.update(node[3][0])
1637
# any node we are querying must be present
1638
present_parents.add(node[1])
1639
unknown_parents = all_parents.difference(present_parents)
1640
present_parents.update(self._present_keys(unknown_parents))
1641
for node in all_nodes:
1643
for parent in node[3][0]:
1644
if parent in present_parents:
1645
parents.append(parent[0])
1646
yield node[1][0], tuple(parents)
1648
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1649
yield node[1][0], ()
1651
def num_versions(self):
1652
return len(list(self._graph_index.iter_all_entries()))
1654
__len__ = num_versions
1656
def get_versions(self):
1657
"""Get all the versions in the file. not topologically sorted."""
1658
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1660
def has_version(self, version_id):
1661
"""True if the version is in the index."""
1662
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1664
def _keys_to_version_ids(self, keys):
1665
return tuple(key[0] for key in keys)
1667
def get_position(self, version_id):
1668
"""Return details needed to access the version.
1670
:return: a tuple (index, data position, size) to hand to the access
1671
logic to get the record.
1673
node = self._get_node(version_id)
1674
bits = node[2][1:].split(' ')
1675
return node[0], int(bits[0]), int(bits[1])
1677
def get_method(self, version_id):
1678
"""Return compression method of specified version."""
1679
if not self._deltas:
1681
return self._parent_compression(self._get_node(version_id)[3][1])
1683
def _parent_compression(self, reference_list):
1684
# use the second reference list to decide if this is delta'd or not.
1685
if len(reference_list):
1690
def _get_node(self, version_id):
1692
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1694
raise RevisionNotPresent(version_id, self)
1696
def get_options(self, version_id):
1697
"""Return a string represention options.
1701
node = self._get_node(version_id)
1702
if not self._deltas:
1703
options = ['fulltext']
1705
options = [self._parent_compression(node[3][1])]
1706
if node[2][0] == 'N':
1707
options.append('no-eol')
1710
def get_parents(self, version_id):
1711
"""Return parents of specified version ignoring ghosts."""
1712
parents = list(self.iter_parents([version_id]))
1715
raise errors.RevisionNotPresent(version_id, self)
1716
return parents[0][1]
1718
def get_parents_with_ghosts(self, version_id):
1719
"""Return parents of specified version with ghosts."""
1720
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1721
check_present=True))
1722
if not self._parents:
1724
return self._keys_to_version_ids(nodes[0][3][0])
1726
def check_versions_present(self, version_ids):
1727
"""Check that all specified versions are present."""
1728
keys = self._version_ids_to_keys(version_ids)
1729
present = self._present_keys(keys)
1730
missing = keys.difference(present)
1732
raise RevisionNotPresent(missing.pop(), self)
1734
def add_version(self, version_id, options, access_memo, parents):
1735
"""Add a version record to the index."""
1736
return self.add_versions(((version_id, options, access_memo, parents),))
1738
def add_versions(self, versions, random_id=False):
1739
"""Add multiple versions to the index.
1741
This function does not insert data into the Immutable GraphIndex
1742
backing the KnitGraphIndex, instead it prepares data for insertion by
1743
the caller and checks that it is safe to insert then calls
1744
self._add_callback with the prepared GraphIndex nodes.
1746
:param versions: a list of tuples:
1747
(version_id, options, pos, size, parents).
1748
:param random_id: If True the ids being added were randomly generated
1749
and no check for existence will be performed.
1751
if not self._add_callback:
1752
raise errors.ReadOnlyError(self)
1753
# we hope there are no repositories with inconsistent parentage
1758
for (version_id, options, access_memo, parents) in versions:
1759
index, pos, size = access_memo
1760
key = (version_id, )
1761
parents = tuple((parent, ) for parent in parents)
1762
if 'no-eol' in options:
1766
value += "%d %d" % (pos, size)
1767
if not self._deltas:
1768
if 'line-delta' in options:
1769
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
1772
if 'line-delta' in options:
1773
node_refs = (parents, (parents[0],))
1775
node_refs = (parents, ())
1777
node_refs = (parents, )
1780
raise KnitCorrupt(self, "attempt to add node with parents "
1781
"in parentless index.")
1783
keys[key] = (value, node_refs)
1785
present_nodes = self._get_entries(keys)
1786
for (index, key, value, node_refs) in present_nodes:
1787
if (value, node_refs) != keys[key]:
1788
raise KnitCorrupt(self, "inconsistent details in add_versions"
1789
": %s %s" % ((value, node_refs), keys[key]))
1793
for key, (value, node_refs) in keys.iteritems():
1794
result.append((key, value, node_refs))
1796
for key, (value, node_refs) in keys.iteritems():
1797
result.append((key, value))
1798
self._add_callback(result)
1800
def _version_ids_to_keys(self, version_ids):
1801
return set((version_id, ) for version_id in version_ids)
1804
class _KnitAccess(object):
1805
"""Access to knit records in a .knit file."""
1807
def __init__(self, transport, filename, _file_mode, _dir_mode,
1808
_need_to_create, _create_parent_dir):
1809
"""Create a _KnitAccess for accessing and inserting data.
1811
:param transport: The transport the .knit is located on.
1812
:param filename: The filename of the .knit.
1814
self._transport = transport
1815
self._filename = filename
1816
self._file_mode = _file_mode
1817
self._dir_mode = _dir_mode
1818
self._need_to_create = _need_to_create
1819
self._create_parent_dir = _create_parent_dir
1821
def add_raw_records(self, sizes, raw_data):
1822
"""Add raw knit bytes to a storage area.
1824
The data is spooled to whereever the access method is storing data.
1826
:param sizes: An iterable containing the size of each raw data segment.
1827
:param raw_data: A bytestring containing the data.
1828
:return: A list of memos to retrieve the record later. Each memo is a
1829
tuple - (index, pos, length), where the index field is always None
1830
for the .knit access method.
1832
assert type(raw_data) == str, \
1833
'data must be plain bytes was %s' % type(raw_data)
1834
if not self._need_to_create:
1835
base = self._transport.append_bytes(self._filename, raw_data)
1837
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1838
create_parent_dir=self._create_parent_dir,
1839
mode=self._file_mode,
1840
dir_mode=self._dir_mode)
1841
self._need_to_create = False
1845
result.append((None, base, size))
1850
"""IFF this data access has its own storage area, initialise it.
1854
self._transport.put_bytes_non_atomic(self._filename, '',
1855
mode=self._file_mode)
1857
def open_file(self):
1858
"""IFF this data access can be represented as a single file, open it.
1860
For knits that are not mapped to a single file on disk this will
1863
:return: None or a file handle.
1866
return self._transport.get(self._filename)
1871
def get_raw_records(self, memos_for_retrieval):
1872
"""Get the raw bytes for a records.
1874
:param memos_for_retrieval: An iterable containing the (index, pos,
1875
length) memo for retrieving the bytes. The .knit method ignores
1876
the index as there is always only a single file.
1877
:return: An iterator over the bytes of the records.
1879
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1880
for pos, data in self._transport.readv(self._filename, read_vector):
1884
class _PackAccess(object):
1885
"""Access to knit records via a collection of packs."""
1887
def __init__(self, index_to_packs, writer=None):
1888
"""Create a _PackAccess object.
1890
:param index_to_packs: A dict mapping index objects to the transport
1891
and file names for obtaining data.
1892
:param writer: A tuple (pack.ContainerWriter, write_index) which
1893
contains the pack to write, and the index that reads from it will
1897
self.container_writer = writer[0]
1898
self.write_index = writer[1]
1900
self.container_writer = None
1901
self.write_index = None
1902
self.indices = index_to_packs
1904
def add_raw_records(self, sizes, raw_data):
1905
"""Add raw knit bytes to a storage area.
1907
The data is spooled to the container writer in one bytes-record per
1910
:param sizes: An iterable containing the size of each raw data segment.
1911
:param raw_data: A bytestring containing the data.
1912
:return: A list of memos to retrieve the record later. Each memo is a
1913
tuple - (index, pos, length), where the index field is the
1914
write_index object supplied to the PackAccess object.
1916
assert type(raw_data) == str, \
1917
'data must be plain bytes was %s' % type(raw_data)
1921
p_offset, p_length = self.container_writer.add_bytes_record(
1922
raw_data[offset:offset+size], [])
1924
result.append((self.write_index, p_offset, p_length))
1928
"""Pack based knits do not get individually created."""
1930
def get_raw_records(self, memos_for_retrieval):
1931
"""Get the raw bytes for a records.
1933
:param memos_for_retrieval: An iterable containing the (index, pos,
1934
length) memo for retrieving the bytes. The Pack access method
1935
looks up the pack to use for a given record in its index_to_pack
1937
:return: An iterator over the bytes of the records.
1939
# first pass, group into same-index requests
1941
current_index = None
1942
for (index, offset, length) in memos_for_retrieval:
1943
if current_index == index:
1944
current_list.append((offset, length))
1946
if current_index is not None:
1947
request_lists.append((current_index, current_list))
1948
current_index = index
1949
current_list = [(offset, length)]
1950
# handle the last entry
1951
if current_index is not None:
1952
request_lists.append((current_index, current_list))
1953
for index, offsets in request_lists:
1954
transport, path = self.indices[index]
1955
reader = pack.make_readv_reader(transport, path, offsets)
1956
for names, read_func in reader.iter_records():
1957
yield read_func(None)
1959
def open_file(self):
1960
"""Pack based knits have no single file."""
1963
def set_writer(self, writer, index, (transport, packname)):
1964
"""Set a writer to use for adding data."""
1965
if index is not None:
1966
self.indices[index] = (transport, packname)
1967
self.container_writer = writer
1968
self.write_index = index
1971
class _KnitData(object):
1972
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1974
The KnitData class provides the logic for parsing and using knit records,
1975
making use of an access method for the low level read and write operations.
1978
def __init__(self, access):
1979
"""Create a KnitData object.
1981
:param access: The access method to use. Access methods such as
1982
_KnitAccess manage the insertion of raw records and the subsequent
1983
retrieval of the same.
1985
self._access = access
1986
self._checked = False
1987
# TODO: jam 20060713 conceptually, this could spill to disk
1988
# if the cached size gets larger than a certain amount
1989
# but it complicates the model a bit, so for now just use
1990
# a simple dictionary
1992
self._do_cache = False
1994
def enable_cache(self):
1995
"""Enable caching of reads."""
1996
self._do_cache = True
1998
def clear_cache(self):
1999
"""Clear the record cache."""
2000
self._do_cache = False
2003
def _open_file(self):
2004
return self._access.open_file()
2006
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2007
"""Convert version_id, digest, lines into a raw data block.
2009
:param dense_lines: The bytes of lines but in a denser form. For
2010
instance, if lines is a list of 1000 bytestrings each ending in \n,
2011
dense_lines may be a list with one line in it, containing all the
2012
1000's lines and their \n's. Using dense_lines if it is already
2013
known is a win because the string join to create bytes in this
2014
function spends less time resizing the final string.
2015
:return: (len, a StringIO instance with the raw data ready to read.)
2017
# Note: using a string copy here increases memory pressure with e.g.
2018
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2019
# when doing the initial commit of a mozilla tree. RBC 20070921
2020
bytes = ''.join(chain(
2021
["version %s %d %s\n" % (version_id,
2024
dense_lines or lines,
2025
["end %s\n" % version_id]))
2026
assert bytes.__class__ == str
2027
compressed_bytes = bytes_to_gzip(bytes)
2028
return len(compressed_bytes), compressed_bytes
2030
def add_raw_records(self, sizes, raw_data):
2031
"""Append a prepared record to the data file.
2033
:param sizes: An iterable containing the size of each raw data segment.
2034
:param raw_data: A bytestring containing the data.
2035
:return: a list of index data for the way the data was stored.
2036
See the access method add_raw_records documentation for more
2039
return self._access.add_raw_records(sizes, raw_data)
2041
def _parse_record_header(self, version_id, raw_data):
2042
"""Parse a record header for consistency.
2044
:return: the header and the decompressor stream.
2045
as (stream, header_record)
2047
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2049
rec = self._check_header(version_id, df.readline())
2050
except Exception, e:
2051
raise KnitCorrupt(self._access,
2052
"While reading {%s} got %s(%s)"
2053
% (version_id, e.__class__.__name__, str(e)))
2056
def _check_header(self, version_id, line):
2059
raise KnitCorrupt(self._access,
2060
'unexpected number of elements in record header')
2061
if rec[1] != version_id:
2062
raise KnitCorrupt(self._access,
2063
'unexpected version, wanted %r, got %r'
2064
% (version_id, rec[1]))
2067
def _parse_record(self, version_id, data):
2069
# 4168 calls in 2880 217 internal
2070
# 4168 calls to _parse_record_header in 2121
2071
# 4168 calls to readlines in 330
2072
df = GzipFile(mode='rb', fileobj=StringIO(data))
2075
record_contents = df.readlines()
2076
except Exception, e:
2077
raise KnitCorrupt(self._access,
2078
"While reading {%s} got %s(%s)"
2079
% (version_id, e.__class__.__name__, str(e)))
2080
header = record_contents.pop(0)
2081
rec = self._check_header(version_id, header)
2083
last_line = record_contents.pop()
2084
if len(record_contents) != int(rec[2]):
2085
raise KnitCorrupt(self._access,
2086
'incorrect number of lines %s != %s'
2088
% (len(record_contents), int(rec[2]),
2090
if last_line != 'end %s\n' % rec[1]:
2091
raise KnitCorrupt(self._access,
2092
'unexpected version end line %r, wanted %r'
2093
% (last_line, version_id))
2095
return record_contents, rec[3]
2097
def read_records_iter_raw(self, records):
2098
"""Read text records from data file and yield raw data.
2100
This unpacks enough of the text record to validate the id is
2101
as expected but thats all.
2103
# setup an iterator of the external records:
2104
# uses readv so nice and fast we hope.
2106
# grab the disk data needed.
2108
# Don't check _cache if it is empty
2109
needed_offsets = [index_memo for version_id, index_memo
2111
if version_id not in self._cache]
2113
needed_offsets = [index_memo for version_id, index_memo
2116
raw_records = self._access.get_raw_records(needed_offsets)
2118
for version_id, index_memo in records:
2119
if version_id in self._cache:
2120
# This data has already been validated
2121
data = self._cache[version_id]
2123
data = raw_records.next()
2125
self._cache[version_id] = data
2127
# validate the header
2128
df, rec = self._parse_record_header(version_id, data)
2130
yield version_id, data
2132
def read_records_iter(self, records):
2133
"""Read text records from data file and yield result.
2135
The result will be returned in whatever is the fastest to read.
2136
Not by the order requested. Also, multiple requests for the same
2137
record will only yield 1 response.
2138
:param records: A list of (version_id, pos, len) entries
2139
:return: Yields (version_id, contents, digest) in the order
2140
read, not the order requested
2146
# Skip records we have alread seen
2147
yielded_records = set()
2148
needed_records = set()
2149
for record in records:
2150
if record[0] in self._cache:
2151
if record[0] in yielded_records:
2153
yielded_records.add(record[0])
2154
data = self._cache[record[0]]
2155
content, digest = self._parse_record(record[0], data)
2156
yield (record[0], content, digest)
2158
needed_records.add(record)
2159
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2161
needed_records = sorted(set(records), key=operator.itemgetter(1))
2163
if not needed_records:
2166
# The transport optimizes the fetching as well
2167
# (ie, reads continuous ranges.)
2168
raw_data = self._access.get_raw_records(
2169
[index_memo for version_id, index_memo in needed_records])
2171
for (version_id, index_memo), data in \
2172
izip(iter(needed_records), raw_data):
2173
content, digest = self._parse_record(version_id, data)
2175
self._cache[version_id] = data
2176
yield version_id, content, digest
2178
def read_records(self, records):
2179
"""Read records into a dictionary."""
2181
for record_id, content, digest in \
2182
self.read_records_iter(records):
2183
components[record_id] = (content, digest)
2187
class InterKnit(InterVersionedFile):
2188
"""Optimised code paths for knit to knit operations."""
2190
_matching_file_from_factory = KnitVersionedFile
2191
_matching_file_to_factory = KnitVersionedFile
2194
def is_compatible(source, target):
2195
"""Be compatible with knits. """
2197
return (isinstance(source, KnitVersionedFile) and
2198
isinstance(target, KnitVersionedFile))
2199
except AttributeError:
2202
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2203
"""See InterVersionedFile.join."""
2204
assert isinstance(self.source, KnitVersionedFile)
2205
assert isinstance(self.target, KnitVersionedFile)
2207
# If the source and target are mismatched w.r.t. annotations vs
2208
# plain, the data needs to be converted accordingly
2209
if self.source.factory.annotated == self.target.factory.annotated:
2211
elif self.source.factory.annotated:
2212
converter = self._anno_to_plain_converter
2214
# We're converting from a plain to an annotated knit. This requires
2215
# building the annotations from scratch. The generic join code
2216
# handles this implicitly so we delegate to it.
2217
return super(InterKnit, self).join(pb, msg, version_ids,
2220
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2224
pb = ui.ui_factory.nested_progress_bar()
2226
version_ids = list(version_ids)
2227
if None in version_ids:
2228
version_ids.remove(None)
2230
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2231
this_versions = set(self.target._index.get_versions())
2232
# XXX: For efficiency we should not look at the whole index,
2233
# we only need to consider the referenced revisions - they
2234
# must all be present, or the method must be full-text.
2235
# TODO, RBC 20070919
2236
needed_versions = self.source_ancestry - this_versions
2238
if not needed_versions:
2240
full_list = topo_sort(self.source.get_graph())
2242
version_list = [i for i in full_list if (not self.target.has_version(i)
2243
and i in needed_versions)]
2247
copy_queue_records = []
2249
for version_id in version_list:
2250
options = self.source._index.get_options(version_id)
2251
parents = self.source._index.get_parents_with_ghosts(version_id)
2252
# check that its will be a consistent copy:
2253
for parent in parents:
2254
# if source has the parent, we must :
2255
# * already have it or
2256
# * have it scheduled already
2257
# otherwise we don't care
2258
assert (self.target.has_version(parent) or
2259
parent in copy_set or
2260
not self.source.has_version(parent))
2261
index_memo = self.source._index.get_position(version_id)
2262
copy_queue_records.append((version_id, index_memo))
2263
copy_queue.append((version_id, options, parents))
2264
copy_set.add(version_id)
2266
# data suck the join:
2268
total = len(version_list)
2271
for (version_id, raw_data), \
2272
(version_id2, options, parents) in \
2273
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2275
assert version_id == version_id2, 'logic error, inconsistent results'
2277
pb.update("Joining knit", count, total)
2279
size, raw_data = converter(raw_data, version_id, options,
2282
size = len(raw_data)
2283
raw_records.append((version_id, options, parents, size))
2284
raw_datum.append(raw_data)
2285
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2290
def _anno_to_plain_converter(self, raw_data, version_id, options,
2292
"""Convert annotated content to plain content."""
2293
data, digest = self.source._data._parse_record(version_id, raw_data)
2294
if 'fulltext' in options:
2295
content = self.source.factory.parse_fulltext(data, version_id)
2296
lines = self.target.factory.lower_fulltext(content)
2298
delta = self.source.factory.parse_line_delta(data, version_id,
2300
lines = self.target.factory.lower_line_delta(delta)
2301
return self.target._data._record_to_data(version_id, digest, lines)
2304
InterVersionedFile.register_optimiser(InterKnit)
2307
class WeaveToKnit(InterVersionedFile):
2308
"""Optimised code paths for weave to knit operations."""
2310
_matching_file_from_factory = bzrlib.weave.WeaveFile
2311
_matching_file_to_factory = KnitVersionedFile
2314
def is_compatible(source, target):
2315
"""Be compatible with weaves to knits."""
2317
return (isinstance(source, bzrlib.weave.Weave) and
2318
isinstance(target, KnitVersionedFile))
2319
except AttributeError:
2322
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2323
"""See InterVersionedFile.join."""
2324
assert isinstance(self.source, bzrlib.weave.Weave)
2325
assert isinstance(self.target, KnitVersionedFile)
2327
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2332
pb = ui.ui_factory.nested_progress_bar()
2334
version_ids = list(version_ids)
2336
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2337
this_versions = set(self.target._index.get_versions())
2338
needed_versions = self.source_ancestry - this_versions
2340
if not needed_versions:
2342
full_list = topo_sort(self.source.get_graph())
2344
version_list = [i for i in full_list if (not self.target.has_version(i)
2345
and i in needed_versions)]
2349
total = len(version_list)
2350
for version_id in version_list:
2351
pb.update("Converting to knit", count, total)
2352
parents = self.source.get_parents(version_id)
2353
# check that its will be a consistent copy:
2354
for parent in parents:
2355
# if source has the parent, we must already have it
2356
assert (self.target.has_version(parent))
2357
self.target.add_lines(
2358
version_id, parents, self.source.get_lines(version_id))
2365
InterVersionedFile.register_optimiser(WeaveToKnit)
2368
# Deprecated, use PatienceSequenceMatcher instead
2369
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2372
def annotate_knit(knit, revision_id):
2373
"""Annotate a knit with no cached annotations.
2375
This implementation is for knits with no cached annotations.
2376
It will work for knits with cached annotations, but this is not
2379
ancestry = knit.get_ancestry(revision_id)
2380
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2382
for candidate in ancestry:
2383
if candidate in annotations:
2385
parents = knit.get_parents(candidate)
2386
if len(parents) == 0:
2388
elif knit._index.get_method(candidate) != 'line-delta':
2391
parent, sha1, noeol, delta = knit.get_delta(candidate)
2392
blocks = KnitContent.get_line_delta_blocks(delta,
2393
fulltext[parents[0]], fulltext[candidate])
2394
annotations[candidate] = list(annotate.reannotate([annotations[p]
2395
for p in parents], fulltext[candidate], candidate, blocks))
2396
return iter(annotations[revision_id])
2400
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2402
from bzrlib._knit_load_data_py import _load_data_py as _load_data
added
removed
bzrlib/knit.py
