1
# Copyright (C) 2005, 2006 by Canonical Ltd
2
# Written by Martin Pool.
3
# Modified by Johan Rydberg <jrydberg@gnu.org>
4
# Modified by Robert Collins <robert.collins@canonical.com>
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
11
# This program is distributed in the hope that it will be useful,
12
# but WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14
# GNU General Public License for more details.
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
"""Knit versionedfile implementation.
22
A knit is a versioned file implementation that supports efficient append only
26
lifeless: the data file is made up of "delta records". each delta record has a delta header
27
that contains; (1) a version id, (2) the size of the delta (in lines), and (3) the digest of
28
the -expanded data- (ie, the delta applied to the parent). the delta also ends with a
29
end-marker; simply "end VERSION"
31
delta can be line or full contents.a
32
... the 8's there are the index number of the annotation.
33
version robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad 7 c7d23b2a5bd6ca00e8e266cec0ec228158ee9f9e
37
8 e.set('executable', 'yes')
39
8 if elt.get('executable') == 'yes':
40
8 ie.executable = True
41
end robertc@robertcollins.net-20051003014215-ee2990904cc4c7ad
45
09:33 < jrydberg> lifeless: each index is made up of a tuple of; version id, options, position, size, parents
46
09:33 < jrydberg> lifeless: the parents are currently dictionary compressed
47
09:33 < jrydberg> lifeless: (meaning it currently does not support ghosts)
48
09:33 < lifeless> right
49
09:33 < jrydberg> lifeless: the position and size is the range in the data file
52
so the index sequence is the dictionary compressed sequence number used
53
in the deltas to provide line annotation
58
# 10:16 < lifeless> make partial index writes safe
59
# 10:16 < lifeless> implement 'knit.check()' like weave.check()
60
# 10:17 < lifeless> record known ghosts so we can detect when they are filled in rather than the current 'reweave
62
# move sha1 out of the content so that join is faster at verifying parents
63
# record content length ?
66
from cStringIO import StringIO
68
from difflib import SequenceMatcher
69
from gzip import GzipFile
72
import bzrlib.errors as errors
73
from bzrlib.errors import FileExists, NoSuchFile, KnitError, \
74
InvalidRevisionId, KnitCorrupt, KnitHeaderError, \
75
RevisionNotPresent, RevisionAlreadyPresent
76
from bzrlib.trace import mutter
77
from bzrlib.osutils import contains_whitespace, contains_linebreaks, \
79
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
80
from bzrlib.tsort import topo_sort
83
# TODO: Split out code specific to this format into an associated object.
85
# TODO: Can we put in some kind of value to check that the index and data
86
# files belong together?
88
# TODO: accomodate binaries, perhaps by storing a byte count
90
# TODO: function to check whole file
92
# TODO: atomically append data, then measure backwards from the cursor
93
# position after writing to work out where it was located. we may need to
94
# bypass python file buffering.
97
INDEX_SUFFIX = '.kndx'
100
class KnitContent(object):
101
"""Content of a knit version to which deltas can be applied."""
103
def __init__(self, lines):
106
def annotate_iter(self):
107
"""Yield tuples of (origin, text) for each content line."""
108
for origin, text in self._lines:
112
"""Return a list of (origin, text) tuples."""
113
return list(self.annotate_iter())
115
def apply_delta(self, delta):
116
"""Apply delta to this content."""
118
for start, end, count, lines in delta:
119
self._lines[offset+start:offset+end] = lines
120
offset = offset + (start - end) + count
122
def line_delta_iter(self, new_lines):
123
"""Generate line-based delta from new_lines to this content."""
124
new_texts = [text for origin, text in new_lines._lines]
125
old_texts = [text for origin, text in self._lines]
126
s = difflib.SequenceMatcher(None, old_texts, new_texts)
127
for op in s.get_opcodes():
130
yield (op[1], op[2], op[4]-op[3], new_lines._lines[op[3]:op[4]])
132
def line_delta(self, new_lines):
133
return list(self.line_delta_iter(new_lines))
136
return [text for origin, text in self._lines]
139
class _KnitFactory(object):
140
"""Base factory for creating content objects."""
142
def make(self, lines, version):
143
num_lines = len(lines)
144
return KnitContent(zip([version] * num_lines, lines))
147
class KnitAnnotateFactory(_KnitFactory):
148
"""Factory for creating annotated Content objects."""
152
def parse_fulltext(self, content, version):
155
origin, text = line.split(' ', 1)
156
lines.append((int(origin), text))
157
return KnitContent(lines)
159
def parse_line_delta_iter(self, lines):
161
header = lines.pop(0)
162
start, end, c = [int(n) for n in header.split(',')]
165
origin, text = lines.pop(0).split(' ', 1)
166
contents.append((int(origin), text))
167
yield start, end, c, contents
169
def parse_line_delta(self, lines, version):
170
return list(self.parse_line_delta_iter(lines))
172
def lower_fulltext(self, content):
173
return ['%d %s' % (o, t) for o, t in content._lines]
175
def lower_line_delta(self, delta):
177
for start, end, c, lines in delta:
178
out.append('%d,%d,%d\n' % (start, end, c))
179
for origin, text in lines:
180
out.append('%d %s' % (origin, text))
184
class KnitPlainFactory(_KnitFactory):
185
"""Factory for creating plain Content objects."""
189
def parse_fulltext(self, content, version):
190
return self.make(content, version)
192
def parse_line_delta_iter(self, lines, version):
194
header = lines.pop(0)
195
start, end, c = [int(n) for n in header.split(',')]
196
yield start, end, c, zip([version] * c, lines[:c])
199
def parse_line_delta(self, lines, version):
200
return list(self.parse_line_delta_iter(lines, version))
202
def lower_fulltext(self, content):
203
return content.text()
205
def lower_line_delta(self, delta):
207
for start, end, c, lines in delta:
208
out.append('%d,%d,%d\n' % (start, end, c))
209
out.extend([text for origin, text in lines])
213
def make_empty_knit(transport, relpath):
214
"""Construct a empty knit at the specified location."""
215
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
219
class KnitVersionedFile(VersionedFile):
220
"""Weave-like structure with faster random access.
222
A knit stores a number of texts and a summary of the relationships
223
between them. Texts are identified by a string version-id. Texts
224
are normally stored and retrieved as a series of lines, but can
225
also be passed as single strings.
227
Lines are stored with the trailing newline (if any) included, to
228
avoid special cases for files with no final newline. Lines are
229
composed of 8-bit characters, not unicode. The combination of
230
these approaches should mean any 'binary' file can be safely
231
stored and retrieved.
234
def __init__(self, relpath, transport, file_mode=None, access_mode=None, factory=None,
235
basis_knit=None, delta=True, create=False):
236
"""Construct a knit at location specified by relpath.
238
:param create: If not True, only open an existing knit.
240
if access_mode is None:
242
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
243
assert not basis_knit or isinstance(basis_knit, KnitVersionedFile), \
246
self.transport = transport
247
self.filename = relpath
248
self.basis_knit = basis_knit
249
self.factory = factory or KnitAnnotateFactory()
250
self.writable = (access_mode == 'w')
253
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
254
access_mode, create=create)
255
self._data = _KnitData(transport, relpath + DATA_SUFFIX,
256
access_mode, create=not len(self.versions()))
258
def copy_to(self, name, transport):
259
"""See VersionedFile.copy_to()."""
260
# copy the current index to a temp index to avoid racing with local
262
transport.put(name + INDEX_SUFFIX + '.tmp', self.transport.get(self._index._filename))
264
transport.put(name + DATA_SUFFIX, self._data._open_file())
265
# rename the copied index into place
266
transport.rename(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
268
def create_empty(self, name, transport, mode=None):
269
return KnitVersionedFile(name, transport, factory=self.factory, delta=self.delta, create=True)
273
"""See VersionedFile.get_suffixes()."""
274
return [DATA_SUFFIX, INDEX_SUFFIX]
277
"""See VersionedFile.versions."""
278
return self._index.get_versions()
280
def has_version(self, version_id):
281
"""See VersionedFile.has_version."""
282
return self._index.has_version(version_id)
284
__contains__ = has_version
286
def _merge_annotations(self, content, parents):
287
"""Merge annotations for content. This is done by comparing
288
the annotations based on changed to the text."""
289
for parent_id in parents:
290
merge_content = self._get_content(parent_id)
291
seq = SequenceMatcher(None, merge_content.text(), content.text())
292
for i, j, n in seq.get_matching_blocks():
295
content._lines[j:j+n] = merge_content._lines[i:i+n]
297
def _get_components(self, version_id):
298
"""Return a list of (version_id, method, data) tuples that
299
makes up version specified by version_id of the knit.
301
The components should be applied in the order of the returned
304
The basis knit will be used to the largest extent possible
305
since it is assumed that accesses to it is faster.
307
# needed_revisions holds a list of (method, version_id) of
308
# versions that is needed to be fetched to construct the final
309
# version of the file.
311
# basis_revisions is a list of versions that needs to be
312
# fetched but exists in the basis knit.
314
basis = self.basis_knit
321
if basis and basis._index.has_version(cursor):
323
basis_versions.append(cursor)
324
method = picked_knit._index.get_method(cursor)
325
needed_versions.append((method, cursor))
326
if method == 'fulltext':
328
cursor = picked_knit.get_parents(cursor)[0]
333
for comp_id in basis_versions:
334
data_pos, data_size = basis._index.get_data_position(comp_id)
335
records.append((piece_id, data_pos, data_size))
336
components.update(basis._data.read_records(records))
339
for comp_id in [vid for method, vid in needed_versions
340
if vid not in basis_versions]:
341
data_pos, data_size = self._index.get_position(comp_id)
342
records.append((comp_id, data_pos, data_size))
343
components.update(self._data.read_records(records))
345
# get_data_records returns a mapping with the version id as
346
# index and the value as data. The order the components need
347
# to be applied is held by needed_versions (reversed).
349
for method, comp_id in reversed(needed_versions):
350
out.append((comp_id, method, components[comp_id]))
354
def _get_content(self, version_id):
355
"""Returns a content object that makes up the specified
357
if not self.has_version(version_id):
358
raise RevisionNotPresent(version_id, self.filename)
360
if self.basis_knit and version_id in self.basis_knit:
361
return self.basis_knit._get_content(version_id)
364
components = self._get_components(version_id)
365
for component_id, method, (data, digest) in components:
366
version_idx = self._index.lookup(component_id)
367
if method == 'fulltext':
368
assert content is None
369
content = self.factory.parse_fulltext(data, version_idx)
370
elif method == 'line-delta':
371
delta = self.factory.parse_line_delta(data, version_idx)
372
content.apply_delta(delta)
374
if 'no-eol' in self._index.get_options(version_id):
375
line = content._lines[-1][1].rstrip('\n')
376
content._lines[-1] = (content._lines[-1][0], line)
378
if sha_strings(content.text()) != digest:
379
raise KnitCorrupt(self.filename, 'sha-1 does not match')
383
def _check_versions_present(self, version_ids):
384
"""Check that all specified versions are present."""
385
version_ids = set(version_ids)
386
for r in list(version_ids):
387
if self._index.has_version(r):
388
version_ids.remove(r)
390
raise RevisionNotPresent(list(version_ids)[0], self.filename)
392
def add_lines(self, version_id, parents, lines):
393
"""See VersionedFile.add_lines."""
394
assert self.writable, "knit is not opened for write"
395
### FIXME escape. RBC 20060228
396
if contains_whitespace(version_id):
397
raise InvalidRevisionId(version_id)
398
if self.has_version(version_id):
399
raise RevisionAlreadyPresent(version_id, self.filename)
401
if False or __debug__:
403
assert '\n' not in l[:-1]
405
self._check_versions_present(parents)
406
return self._add(version_id, lines[:], parents, self.delta)
408
def _add(self, version_id, lines, parents, delta):
409
"""Add a set of lines on top of version specified by parents.
411
If delta is true, compress the text as a line-delta against
414
if delta and not parents:
417
digest = sha_strings(lines)
420
if lines[-1][-1] != '\n':
421
options.append('no-eol')
422
lines[-1] = lines[-1] + '\n'
424
lines = self.factory.make(lines, len(self._index))
425
if self.factory.annotated and len(parents) > 0:
426
# Merge annotations from parent texts if so is needed.
427
self._merge_annotations(lines, parents)
429
if parents and delta:
430
# To speed the extract of texts the delta chain is limited
431
# to a fixed number of deltas. This should minimize both
432
# I/O and the time spend applying deltas.
434
delta_parents = parents
436
parent = delta_parents[0]
437
method = self._index.get_method(parent)
438
if method == 'fulltext':
440
delta_parents = self._index.get_parents(parent)
442
if method == 'line-delta':
446
options.append('line-delta')
447
content = self._get_content(parents[0])
448
delta_hunks = content.line_delta(lines)
449
store_lines = self.factory.lower_line_delta(delta_hunks)
451
options.append('fulltext')
452
store_lines = self.factory.lower_fulltext(lines)
454
where, size = self._data.add_record(version_id, digest, store_lines)
455
self._index.add_version(version_id, options, where, size, parents)
457
def check(self, progress_bar=None):
458
"""See VersionedFile.check()."""
460
def clone_text(self, new_version_id, old_version_id, parents):
461
"""See VersionedFile.clone_text()."""
462
# FIXME RBC 20060228 make fast by only inserting an index with null delta.
463
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
465
def get_lines(self, version_id):
466
"""See VersionedFile.get_lines()."""
467
return self._get_content(version_id).text()
469
def iter_lines_added_or_present_in_versions(self, version_ids=None):
470
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
471
if version_ids is None:
472
version_ids = self.versions()
473
# we dont care about inclusions, the caller cares.
474
# but we need to setup a list of records to visit.
475
# we need version_id, position, length
476
version_id_records = []
477
for version_id in version_ids:
478
if not self.has_version(version_id):
479
raise RevisionNotPresent(version_id, self.filename)
480
data_pos, length = self._index.get_position(version_id)
481
version_id_records.append((version_id, data_pos, length))
482
for version_id, data, sha_value in \
483
self._data.read_records_iter(version_id_records):
484
method = self._index.get_method(version_id)
485
version_idx = self._index.lookup(version_id)
486
assert method in ('fulltext', 'line-delta')
487
if method == 'fulltext':
488
content = self.factory.parse_fulltext(data, version_idx)
489
for line in content.text():
492
delta = self.factory.parse_line_delta(data, version_idx)
493
for start, end, count, lines in delta:
494
for origin, line in lines:
497
def num_versions(self):
498
"""See VersionedFile.num_versions()."""
499
return self._index.num_versions()
501
__len__ = num_versions
503
def annotate_iter(self, version_id):
504
"""See VersionedFile.annotate_iter."""
505
content = self._get_content(version_id)
506
for origin, text in content.annotate_iter():
507
yield self._index.idx_to_name(origin), text
509
def get_parents(self, version_id):
510
"""See VersionedFile.get_parents."""
511
self._check_versions_present([version_id])
512
return list(self._index.get_parents(version_id))
514
def get_ancestry(self, versions):
515
"""See VersionedFile.get_ancestry."""
516
if isinstance(versions, basestring):
517
versions = [versions]
520
self._check_versions_present(versions)
521
return self._index.get_ancestry(versions)
523
def _reannotate_line_delta(self, other, lines, new_version_id,
525
"""Re-annotate line-delta and return new delta."""
527
for start, end, count, contents \
528
in self.factory.parse_line_delta_iter(lines):
530
for origin, line in contents:
531
old_version_id = other._index.idx_to_name(origin)
532
if old_version_id == new_version_id:
533
idx = new_version_idx
535
idx = self._index.lookup(old_version_id)
536
new_lines.append((idx, line))
537
new_delta.append((start, end, count, new_lines))
539
return self.factory.lower_line_delta(new_delta)
541
def _reannotate_fulltext(self, other, lines, new_version_id,
543
"""Re-annotate fulltext and return new version."""
544
content = self.factory.parse_fulltext(lines, new_version_idx)
546
for origin, line in content.annotate_iter():
547
old_version_id = other._index.idx_to_name(origin)
548
if old_version_id == new_version_id:
549
idx = new_version_idx
551
idx = self._index.lookup(old_version_id)
552
new_lines.append((idx, line))
554
return self.factory.lower_fulltext(KnitContent(new_lines))
556
#@deprecated_method(zero_eight)
557
def walk(self, version_ids):
558
"""See VersionedFile.walk."""
559
# We take the short path here, and extract all relevant texts
560
# and put them in a weave and let that do all the work. Far
561
# from optimal, but is much simpler.
562
# FIXME RB 20060228 this really is inefficient!
563
from bzrlib.weave import Weave
565
w = Weave(self.filename)
566
ancestry = self.get_ancestry(version_ids)
567
sorted_graph = topo_sort(self._index.get_graph())
568
version_list = [vid for vid in sorted_graph if vid in ancestry]
570
for version_id in version_list:
571
lines = self.get_lines(version_id)
572
w.add_lines(version_id, self.get_parents(version_id), lines)
574
for lineno, insert_id, dset, line in w.walk(version_ids):
575
yield lineno, insert_id, dset, line
578
class _KnitComponentFile(object):
579
"""One of the files used to implement a knit database"""
581
def __init__(self, transport, filename, mode):
582
self._transport = transport
583
self._filename = filename
586
def write_header(self):
587
old_len = self._transport.append(self._filename, StringIO(self.HEADER))
589
raise KnitCorrupt(self._filename, 'misaligned after writing header')
591
def check_header(self, fp):
592
line = fp.read(len(self.HEADER))
593
if line != self.HEADER:
594
raise KnitHeaderError(badline=line)
597
"""Commit is a nop."""
600
return '%s(%s)' % (self.__class__.__name__, self._filename)
603
class _KnitIndex(_KnitComponentFile):
604
"""Manages knit index file.
606
The index is already kept in memory and read on startup, to enable
607
fast lookups of revision information. The cursor of the index
608
file is always pointing to the end, making it easy to append
611
_cache is a cache for fast mapping from version id to a Index
614
_history is a cache for fast mapping from indexes to version ids.
616
The index data format is dictionary compressed when it comes to
617
parent references; a index entry may only have parents that with a
618
lover index number. As a result, the index is topological sorted.
620
Duplicate entries may be written to the index for a single version id
621
if this is done then the latter one completely replaces the former:
622
this allows updates to correct version and parent information.
623
Note that the two entries may share the delta, and that successive
624
annotations and references MUST point to the first entry.
627
HEADER = "# bzr knit index 7\n"
629
def _cache_version(self, version_id, options, pos, size, parents):
630
val = (version_id, options, pos, size, parents)
631
self._cache[version_id] = val
632
if not version_id in self._history:
633
self._history.append(version_id)
635
def _iter_index(self, fp):
637
for l in lines.splitlines(False):
640
def __init__(self, transport, filename, mode, create=False):
641
_KnitComponentFile.__init__(self, transport, filename, mode)
643
# position in _history is the 'official' index for a revision
644
# but the values may have come from a newer entry.
645
# so - wc -l of a knit index is != the number of uniqe names
649
fp = self._transport.get(self._filename)
650
self.check_header(fp)
651
for rec in self._iter_index(fp):
652
self._cache_version(rec[0], rec[1].split(','), int(rec[2]), int(rec[3]),
653
[self._history[int(i)] for i in rec[4:]])
654
except NoSuchFile, e:
655
if mode != 'w' or not create:
661
for version_id, index in self._cache.iteritems():
662
graph.append((version_id, index[4]))
665
def get_ancestry(self, versions):
666
"""See VersionedFile.get_ancestry."""
667
# get a graph of all the mentioned versions:
669
pending = set(versions)
671
version = pending.pop()
672
parents = self._cache[version][4]
673
for parent in parents:
674
if parent not in graph:
676
graph[version] = parents
677
return topo_sort(graph.items())
679
def num_versions(self):
680
return len(self._history)
682
__len__ = num_versions
684
def get_versions(self):
687
def idx_to_name(self, idx):
688
return self._history[idx]
690
def lookup(self, version_id):
691
assert version_id in self._cache
692
return self._history.index(version_id)
694
def add_version(self, version_id, options, pos, size, parents):
695
"""Add a version record to the index."""
696
self._cache_version(version_id, options, pos, size, parents)
698
content = "%s %s %s %s %s\n" % (version_id,
702
' '.join([str(self.lookup(vid)) for
704
self._transport.append(self._filename, StringIO(content))
706
def has_version(self, version_id):
707
"""True if the version is in the index."""
708
return self._cache.has_key(version_id)
710
def get_position(self, version_id):
711
"""Return data position and size of specified version."""
712
return (self._cache[version_id][2], \
713
self._cache[version_id][3])
715
def get_method(self, version_id):
716
"""Return compression method of specified version."""
717
options = self._cache[version_id][1]
718
if 'fulltext' in options:
721
assert 'line-delta' in options
724
def get_options(self, version_id):
725
return self._cache[version_id][1]
727
def get_parents(self, version_id):
728
"""Return parents of specified version."""
729
return self._cache[version_id][4]
731
def check_versions_present(self, version_ids):
732
"""Check that all specified versions are present."""
733
version_ids = set(version_ids)
734
for version_id in list(version_ids):
735
if version_id in self._cache:
736
version_ids.remove(version_id)
738
raise RevisionNotPresent(list(version_ids)[0], self.filename)
741
class _KnitData(_KnitComponentFile):
742
"""Contents of the knit data file"""
744
HEADER = "# bzr knit data 7\n"
746
def __init__(self, transport, filename, mode, create=False):
747
_KnitComponentFile.__init__(self, transport, filename, mode)
749
self._checked = False
751
self._transport.put(self._filename, StringIO(''))
753
def _open_file(self):
754
if self._file is None:
756
self._file = self._transport.get(self._filename)
761
def add_record(self, version_id, digest, lines):
762
"""Write new text record to disk. Returns the position in the
763
file where it was written."""
765
data_file = GzipFile(None, mode='wb', fileobj=sio)
766
print >>data_file, "version %s %d %s" % (version_id, len(lines), digest)
767
data_file.writelines(lines)
768
print >>data_file, "end %s\n" % version_id
771
content = sio.getvalue()
772
start_pos = self._transport.append(self._filename, StringIO(content))
773
return start_pos, len(content)
775
def _parse_record(self, version_id, data):
776
df = GzipFile(mode='rb', fileobj=StringIO(data))
777
rec = df.readline().split()
779
raise KnitCorrupt(self._filename, 'unexpected number of records')
780
if rec[1] != version_id:
781
raise KnitCorrupt(self.file.name,
782
'unexpected version, wanted %r' % version_id)
784
record_contents = self._read_record_contents(df, lines)
786
if l != 'end %s\n' % version_id:
787
raise KnitCorrupt(self._filename, 'unexpected version end line %r, wanted %r'
789
return record_contents, rec[3]
791
def _read_record_contents(self, df, record_lines):
792
"""Read and return n lines from datafile."""
794
for i in range(record_lines):
795
r.append(df.readline())
798
def read_records_iter(self, records):
799
"""Read text records from data file and yield result.
801
Each passed record is a tuple of (version_id, pos, len) and
802
will be read in the given order. Yields (version_id,
806
class ContinuousRange:
807
def __init__(self, rec_id, pos, size):
809
self.end_pos = pos + size
810
self.versions = [(rec_id, pos, size)]
812
def add(self, rec_id, pos, size):
813
if self.end_pos != pos:
815
self.end_pos = pos + size
816
self.versions.append((rec_id, pos, size))
820
for rec_id, pos, size in self.versions:
821
yield rec_id, fp.read(size)
823
# We take it that the transport optimizes the fetching as good
824
# as possible (ie, reads continous ranges.)
825
response = self._transport.readv(self._filename,
826
[(pos, size) for version_id, pos, size in records])
828
for (record_id, pos, size), (pos, data) in zip(records, response):
829
content, digest = self._parse_record(record_id, data)
830
yield record_id, content, digest
832
def read_records(self, records):
833
"""Read records into a dictionary."""
835
for record_id, content, digest in self.read_records_iter(records):
836
components[record_id] = (content, digest)
840
class InterKnit(InterVersionedFile):
841
"""Optimised code paths for knit to knit operations."""
843
_matching_file_factory = KnitVersionedFile
846
def is_compatible(source, target):
847
"""Be compatible with knits. """
849
return (isinstance(source, KnitVersionedFile) and
850
isinstance(target, KnitVersionedFile))
851
except AttributeError:
854
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
855
"""See InterVersionedFile.join."""
856
assert isinstance(self.source, KnitVersionedFile)
857
assert isinstance(self.target, KnitVersionedFile)
859
if version_ids is None:
860
version_ids = self.source.versions()
862
if not ignore_missing:
863
self.source._check_versions_present(version_ids)
865
version_ids = set(self.source.versions()).intersection(
872
from bzrlib.progress import DummyProgress
875
version_ids = list(version_ids)
876
if None in version_ids:
877
version_ids.remove(None)
879
self.source_ancestry = set(self.source.get_ancestry(version_ids))
880
this_versions = set(self.target._index.get_versions())
881
needed_versions = self.source_ancestry - this_versions
882
cross_check_versions = self.source_ancestry.intersection(this_versions)
883
mismatched_versions = set()
884
for version in cross_check_versions:
885
# scan to include needed parents.
886
n1 = set(self.target.get_parents(version))
887
n2 = set(self.source.get_parents(version))
889
# FIXME TEST this check for cycles being introduced works
890
# the logic is we have a cycle if in our graph we are an
891
# ancestor of any of the n2 revisions.
897
parent_ancestors = self.source.get_ancestry(parent)
898
if version in parent_ancestors:
899
raise errors.GraphCycleError([parent, version])
900
# ensure this parent will be available later.
901
new_parents = n2.difference(n1)
902
needed_versions.update(new_parents.difference(this_versions))
903
mismatched_versions.add(version)
905
if not needed_versions and not cross_check_versions:
907
full_list = topo_sort(self.source._index.get_graph())
909
version_list = [i for i in full_list if (not self.target.has_version(i)
910
and i in needed_versions)]
913
for version_id in version_list:
914
data_pos, data_size = self.source._index.get_position(version_id)
915
records.append((version_id, data_pos, data_size))
918
for version_id, lines, digest \
919
in self.source._data.read_records_iter(records):
920
options = self.source._index.get_options(version_id)
921
parents = self.source._index.get_parents(version_id)
923
for parent in parents:
924
assert self.target.has_version(parent)
926
if self.target.factory.annotated:
927
# FIXME jrydberg: it should be possible to skip
928
# re-annotating components if we know that we are
929
# going to pull all revisions in the same order.
930
new_version_id = version_id
931
new_version_idx = self.target._index.num_versions()
932
if 'fulltext' in options:
933
lines = self.target._reannotate_fulltext(self.source, lines,
934
new_version_id, new_version_idx)
935
elif 'line-delta' in options:
936
lines = self.target._reannotate_line_delta(self.source, lines,
937
new_version_id, new_version_idx)
940
pb.update("Joining knit", count, len(version_list))
942
pos, size = self.target._data.add_record(version_id, digest, lines)
943
self.target._index.add_version(version_id, options, pos, size, parents)
945
for version in mismatched_versions:
946
n1 = set(self.target.get_parents(version))
947
n2 = set(self.source.get_parents(version))
948
# write a combined record to our history.
949
new_parents = self.target.get_parents(version) + list(n2.difference(n1))
950
current_values = self.target._index._cache[version]
951
self.target._index.add_version(version,
960
InterVersionedFile.register_optimiser(InterKnit)
added
removed
bzrlib/knit.py
