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):
236
"""Construct a knit at location specified by relpath."""
237
if access_mode is None:
239
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
240
assert not basis_knit or isinstance(basis_knit, KnitVersionedFile), \
243
self.transport = transport
244
self.filename = relpath
245
self.basis_knit = basis_knit
246
self.factory = factory or KnitAnnotateFactory()
247
self.writable = (access_mode == 'w')
250
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
252
self._data = _KnitData(transport, relpath + DATA_SUFFIX,
255
def copy_to(self, name, transport):
256
"""See VersionedFile.copy_to()."""
257
# copy the current index to a temp index to avoid racing with local
259
transport.put(name + INDEX_SUFFIX + '.tmp', self.transport.get(self._index._filename))
261
transport.put(name + DATA_SUFFIX, self._data._open_file())
262
# rename the copied index into place
263
transport.rename(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
265
def create_empty(self, name, transport, mode=None):
266
return KnitVersionedFile(name, transport, factory=self.factory, delta=self.delta)
270
"""See VersionedFile.get_suffixes()."""
271
return [DATA_SUFFIX, INDEX_SUFFIX]
274
"""See VersionedFile.versions."""
275
return self._index.get_versions()
277
def has_version(self, version_id):
278
"""See VersionedFile.has_version."""
279
return self._index.has_version(version_id)
281
__contains__ = has_version
283
def _merge_annotations(self, content, parents):
284
"""Merge annotations for content. This is done by comparing
285
the annotations based on changed to the text."""
286
for parent_id in parents:
287
merge_content = self._get_content(parent_id)
288
seq = SequenceMatcher(None, merge_content.text(), content.text())
289
for i, j, n in seq.get_matching_blocks():
292
content._lines[j:j+n] = merge_content._lines[i:i+n]
294
def _get_components(self, version_id):
295
"""Return a list of (version_id, method, data) tuples that
296
makes up version specified by version_id of the knit.
298
The components should be applied in the order of the returned
301
The basis knit will be used to the largest extent possible
302
since it is assumed that accesses to it is faster.
304
# needed_revisions holds a list of (method, version_id) of
305
# versions that is needed to be fetched to construct the final
306
# version of the file.
308
# basis_revisions is a list of versions that needs to be
309
# fetched but exists in the basis knit.
311
basis = self.basis_knit
318
if basis and basis._index.has_version(cursor):
320
basis_versions.append(cursor)
321
method = picked_knit._index.get_method(cursor)
322
needed_versions.append((method, cursor))
323
if method == 'fulltext':
325
cursor = picked_knit.get_parents(cursor)[0]
330
for comp_id in basis_versions:
331
data_pos, data_size = basis._index.get_data_position(comp_id)
332
records.append((piece_id, data_pos, data_size))
333
components.update(basis._data.read_records(records))
336
for comp_id in [vid for method, vid in needed_versions
337
if vid not in basis_versions]:
338
data_pos, data_size = self._index.get_position(comp_id)
339
records.append((comp_id, data_pos, data_size))
340
components.update(self._data.read_records(records))
342
# get_data_records returns a mapping with the version id as
343
# index and the value as data. The order the components need
344
# to be applied is held by needed_versions (reversed).
346
for method, comp_id in reversed(needed_versions):
347
out.append((comp_id, method, components[comp_id]))
351
def _get_content(self, version_id):
352
"""Returns a content object that makes up the specified
354
if not self.has_version(version_id):
355
raise RevisionNotPresent(version_id, self.filename)
357
if self.basis_knit and version_id in self.basis_knit:
358
return self.basis_knit._get_content(version_id)
361
components = self._get_components(version_id)
362
for component_id, method, (data, digest) in components:
363
version_idx = self._index.lookup(component_id)
364
if method == 'fulltext':
365
assert content is None
366
content = self.factory.parse_fulltext(data, version_idx)
367
elif method == 'line-delta':
368
delta = self.factory.parse_line_delta(data, version_idx)
369
content.apply_delta(delta)
371
if 'no-eol' in self._index.get_options(version_id):
372
line = content._lines[-1][1].rstrip('\n')
373
content._lines[-1] = (content._lines[-1][0], line)
375
if sha_strings(content.text()) != digest:
376
raise KnitCorrupt(self.filename, 'sha-1 does not match')
380
def _check_versions_present(self, version_ids):
381
"""Check that all specified versions are present."""
382
version_ids = set(version_ids)
383
for r in list(version_ids):
384
if self._index.has_version(r):
385
version_ids.remove(r)
387
raise RevisionNotPresent(list(version_ids)[0], self.filename)
389
def add_lines(self, version_id, parents, lines):
390
"""See VersionedFile.add_lines."""
391
assert self.writable, "knit is not opened for write"
392
### FIXME escape. RBC 20060228
393
if contains_whitespace(version_id):
394
raise InvalidRevisionId(version_id)
395
if self.has_version(version_id):
396
raise RevisionAlreadyPresent(version_id, self.filename)
398
if True or __debug__:
400
assert '\n' not in l[:-1]
402
self._check_versions_present(parents)
403
return self._add(version_id, lines[:], parents, self.delta)
405
def _add(self, version_id, lines, parents, delta):
406
"""Add a set of lines on top of version specified by parents.
408
If delta is true, compress the text as a line-delta against
411
if delta and not parents:
414
digest = sha_strings(lines)
417
if lines[-1][-1] != '\n':
418
options.append('no-eol')
419
lines[-1] = lines[-1] + '\n'
421
lines = self.factory.make(lines, len(self._index))
422
if self.factory.annotated and len(parents) > 0:
423
# Merge annotations from parent texts if so is needed.
424
self._merge_annotations(lines, parents)
426
if parents and delta:
427
# To speed the extract of texts the delta chain is limited
428
# to a fixed number of deltas. This should minimize both
429
# I/O and the time spend applying deltas.
431
delta_parents = parents
433
parent = delta_parents[0]
434
method = self._index.get_method(parent)
435
if method == 'fulltext':
437
delta_parents = self._index.get_parents(parent)
439
if method == 'line-delta':
443
options.append('line-delta')
444
content = self._get_content(parents[0])
445
delta_hunks = content.line_delta(lines)
446
store_lines = self.factory.lower_line_delta(delta_hunks)
448
options.append('fulltext')
449
store_lines = self.factory.lower_fulltext(lines)
451
where, size = self._data.add_record(version_id, digest, store_lines)
452
self._index.add_version(version_id, options, where, size, parents)
454
def check(self, progress_bar=None):
455
"""See VersionedFile.check()."""
457
def clone_text(self, new_version_id, old_version_id, parents):
458
"""See VersionedFile.clone_text()."""
459
# FIXME RBC 20060228 make fast by only inserting an index with null delta.
460
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
462
def get_lines(self, version_id):
463
"""See VersionedFile.get_lines()."""
464
return self._get_content(version_id).text()
466
def num_versions(self):
467
"""See VersionedFile.num_versions()."""
468
return self._index.num_versions()
470
__len__ = num_versions
472
def annotate_iter(self, version_id):
473
"""See VersionedFile.annotate_iter."""
474
content = self._get_content(version_id)
475
for origin, text in content.annotate_iter():
476
yield self._index.idx_to_name(origin), text
478
def get_parents(self, version_id):
479
"""See VersionedFile.get_parents."""
480
self._check_versions_present([version_id])
481
return list(self._index.get_parents(version_id))
483
def get_ancestry(self, versions):
484
"""See VersionedFile.get_ancestry."""
485
if isinstance(versions, basestring):
486
versions = [versions]
489
self._check_versions_present(versions)
490
return self._index.get_ancestry(versions)
492
def _reannotate_line_delta(self, other, lines, new_version_id,
494
"""Re-annotate line-delta and return new delta."""
496
for start, end, count, contents \
497
in self.factory.parse_line_delta_iter(lines):
499
for origin, line in contents:
500
old_version_id = other._index.idx_to_name(origin)
501
if old_version_id == new_version_id:
502
idx = new_version_idx
504
idx = self._index.lookup(old_version_id)
505
new_lines.append((idx, line))
506
new_delta.append((start, end, count, new_lines))
508
return self.factory.lower_line_delta(new_delta)
510
def _reannotate_fulltext(self, other, lines, new_version_id,
512
"""Re-annotate fulltext and return new version."""
513
content = self.factory.parse_fulltext(lines, new_version_idx)
515
for origin, line in content.annotate_iter():
516
old_version_id = other._index.idx_to_name(origin)
517
if old_version_id == new_version_id:
518
idx = new_version_idx
520
idx = self._index.lookup(old_version_id)
521
new_lines.append((idx, line))
523
return self.factory.lower_fulltext(KnitContent(new_lines))
525
def walk(self, version_ids):
526
"""See VersionedFile.walk."""
527
# We take the short path here, and extract all relevant texts
528
# and put them in a weave and let that do all the work. Far
529
# from optimal, but is much simpler.
530
# FIXME RB 20060228 this really is inefficient!
531
from bzrlib.weave import Weave
533
w = Weave(self.filename)
534
ancestry = self.get_ancestry(version_ids)
535
sorted_graph = topo_sort(self._index.get_graph())
536
version_list = [vid for vid in sorted_graph if vid in ancestry]
538
for version_id in version_list:
539
lines = self.get_lines(version_id)
540
w.add_lines(version_id, self.get_parents(version_id), lines)
542
for lineno, insert_id, dset, line in w.walk(version_ids):
543
yield lineno, insert_id, dset, line
546
class _KnitComponentFile(object):
547
"""One of the files used to implement a knit database"""
549
def __init__(self, transport, filename, mode):
550
self._transport = transport
551
self._filename = filename
554
def write_header(self):
555
old_len = self._transport.append(self._filename, StringIO(self.HEADER))
557
raise KnitCorrupt(self._filename, 'misaligned after writing header')
559
def check_header(self, fp):
560
line = fp.read(len(self.HEADER))
561
if line != self.HEADER:
562
raise KnitHeaderError(badline=line)
565
"""Commit is a nop."""
568
return '%s(%s)' % (self.__class__.__name__, self._filename)
571
class _KnitIndex(_KnitComponentFile):
572
"""Manages knit index file.
574
The index is already kept in memory and read on startup, to enable
575
fast lookups of revision information. The cursor of the index
576
file is always pointing to the end, making it easy to append
579
_cache is a cache for fast mapping from version id to a Index
582
_history is a cache for fast mapping from indexes to version ids.
584
The index data format is dictionary compressed when it comes to
585
parent references; a index entry may only have parents that with a
586
lover index number. As a result, the index is topological sorted.
588
Duplicate entries may be written to the index for a single version id
589
if this is done then the latter one completely replaces the former:
590
this allows updates to correct version and parent information.
591
Note that the two entries may share the delta, and that successive
592
annotations and references MUST point to the first entry.
595
HEADER = "# bzr knit index 7\n"
597
def _cache_version(self, version_id, options, pos, size, parents):
598
val = (version_id, options, pos, size, parents)
599
self._cache[version_id] = val
600
if not version_id in self._history:
601
self._history.append(version_id)
603
def _iter_index(self, fp):
605
for l in lines.splitlines(False):
608
def __init__(self, transport, filename, mode):
609
_KnitComponentFile.__init__(self, transport, filename, mode)
611
# position in _history is the 'official' index for a revision
612
# but the values may have come from a newer entry.
613
# so - wc -l of a knit index is != the number of uniqe names
617
fp = self._transport.get(self._filename)
618
self.check_header(fp)
619
for rec in self._iter_index(fp):
620
self._cache_version(rec[0], rec[1].split(','), int(rec[2]), int(rec[3]),
621
[self._history[int(i)] for i in rec[4:]])
622
except NoSuchFile, e:
629
for version_id, index in self._cache.iteritems():
630
graph.append((version_id, index[4]))
633
def get_ancestry(self, versions):
634
"""See VersionedFile.get_ancestry."""
636
for version_id in versions:
637
version_idxs.append(self._history.index(version_id))
639
for v in xrange(max(version_idxs), 0, -1):
640
if self._history[v] in i:
641
# include all its parents
642
i.update(self._cache[self._history[v]][4])
645
def num_versions(self):
646
return len(self._history)
648
__len__ = num_versions
650
def get_versions(self):
653
def idx_to_name(self, idx):
654
return self._history[idx]
656
def lookup(self, version_id):
657
assert version_id in self._cache
658
return self._history.index(version_id)
660
def add_version(self, version_id, options, pos, size, parents):
661
"""Add a version record to the index."""
662
self._cache_version(version_id, options, pos, size, parents)
664
content = "%s %s %s %s %s\n" % (version_id,
668
' '.join([str(self.lookup(vid)) for
670
self._transport.append(self._filename, StringIO(content))
672
def has_version(self, version_id):
673
"""True if the version is in the index."""
674
return self._cache.has_key(version_id)
676
def get_position(self, version_id):
677
"""Return data position and size of specified version."""
678
return (self._cache[version_id][2], \
679
self._cache[version_id][3])
681
def get_method(self, version_id):
682
"""Return compression method of specified version."""
683
options = self._cache[version_id][1]
684
if 'fulltext' in options:
687
assert 'line-delta' in options
690
def get_options(self, version_id):
691
return self._cache[version_id][1]
693
def get_parents(self, version_id):
694
"""Return parents of specified version."""
695
return self._cache[version_id][4]
697
def check_versions_present(self, version_ids):
698
"""Check that all specified versions are present."""
699
version_ids = set(version_ids)
700
for version_id in list(version_ids):
701
if version_id in self._cache:
702
version_ids.remove(version_id)
704
raise RevisionNotPresent(list(version_ids)[0], self.filename)
707
class _KnitData(_KnitComponentFile):
708
"""Contents of the knit data file"""
710
HEADER = "# bzr knit data 7\n"
712
def __init__(self, transport, filename, mode):
713
_KnitComponentFile.__init__(self, transport, filename, mode)
715
self._checked = False
717
def _open_file(self):
718
if self._file is None:
720
self._file = self._transport.get(self._filename)
725
def add_record(self, version_id, digest, lines):
726
"""Write new text record to disk. Returns the position in the
727
file where it was written."""
729
data_file = GzipFile(None, mode='wb', fileobj=sio)
730
print >>data_file, "version %s %d %s" % (version_id, len(lines), digest)
731
data_file.writelines(lines)
732
print >>data_file, "end %s\n" % version_id
735
content = sio.getvalue()
736
start_pos = self._transport.append(self._filename, StringIO(content))
737
return start_pos, len(content)
739
def _parse_record(self, version_id, data):
740
df = GzipFile(mode='rb', fileobj=StringIO(data))
741
rec = df.readline().split()
743
raise KnitCorrupt(self._filename, 'unexpected number of records')
744
if rec[1] != version_id:
745
raise KnitCorrupt(self.file.name,
746
'unexpected version, wanted %r' % version_id)
748
record_contents = self._read_record_contents(df, lines)
750
if l != 'end %s\n' % version_id:
751
raise KnitCorrupt(self._filename, 'unexpected version end line %r, wanted %r'
753
return record_contents, rec[3]
755
def _read_record_contents(self, df, record_lines):
756
"""Read and return n lines from datafile."""
758
for i in range(record_lines):
759
r.append(df.readline())
762
def read_records_iter(self, records):
763
"""Read text records from data file and yield result.
765
Each passed record is a tuple of (version_id, pos, len) and
766
will be read in the given order. Yields (version_id,
770
class ContinuousRange:
771
def __init__(self, rec_id, pos, size):
773
self.end_pos = pos + size
774
self.versions = [(rec_id, pos, size)]
776
def add(self, rec_id, pos, size):
777
if self.end_pos != pos:
779
self.end_pos = pos + size
780
self.versions.append((rec_id, pos, size))
784
for rec_id, pos, size in self.versions:
785
yield rec_id, fp.read(size)
787
fp = self._open_file()
789
# Loop through all records and try to collect as large
790
# continuous region as possible to read.
792
record_id, pos, size = records.pop(0)
793
continuous_range = ContinuousRange(record_id, pos, size)
795
record_id, pos, size = records[0]
796
if continuous_range.add(record_id, pos, size):
800
fp.seek(continuous_range.start_pos, 0)
801
for record_id, data in continuous_range.split(fp):
802
content, digest = self._parse_record(record_id, data)
803
yield record_id, content, digest
807
def read_records(self, records):
808
"""Read records into a dictionary."""
810
for record_id, content, digest in self.read_records_iter(records):
811
components[record_id] = (content, digest)
815
class InterKnit(InterVersionedFile):
816
"""Optimised code paths for knit to knit operations."""
818
_matching_file_factory = KnitVersionedFile
821
def is_compatible(source, target):
822
"""Be compatible with knits. """
824
return (isinstance(source, KnitVersionedFile) and
825
isinstance(target, KnitVersionedFile))
826
except AttributeError:
829
def join(self, pb=None, msg=None, version_ids=None):
830
"""See InterVersionedFile.join."""
831
assert isinstance(self.source, KnitVersionedFile)
832
assert isinstance(self.target, KnitVersionedFile)
834
if version_ids is None:
835
version_ids = self.source.versions()
840
from bzrlib.progress import DummyProgress
843
version_ids = list(version_ids)
844
if None in version_ids:
845
version_ids.remove(None)
847
self.source_ancestry = set(self.source.get_ancestry(version_ids))
848
this_versions = set(self.target._index.get_versions())
849
needed_versions = self.source_ancestry - this_versions
850
cross_check_versions = self.source_ancestry.intersection(this_versions)
851
mismatched_versions = set()
852
for version in cross_check_versions:
853
# scan to include needed parents.
854
n1 = set(self.target.get_parents(version))
855
n2 = set(self.source.get_parents(version))
857
# FIXME TEST this check for cycles being introduced works
858
# the logic is we have a cycle if in our graph we are an
859
# ancestor of any of the n2 revisions.
865
parent_ancestors = self.source.get_ancestry(parent)
866
if version in parent_ancestors:
867
raise errors.GraphCycleError([parent, version])
868
# ensure this parent will be available later.
869
new_parents = n2.difference(n1)
870
needed_versions.update(new_parents.difference(this_versions))
871
mismatched_versions.add(version)
873
if not needed_versions and not cross_check_versions:
875
full_list = topo_sort(self.source._index.get_graph())
877
version_list = [i for i in full_list if (not self.target.has_version(i)
878
and i in needed_versions)]
881
for version_id in version_list:
882
data_pos, data_size = self.source._index.get_position(version_id)
883
records.append((version_id, data_pos, data_size))
886
for version_id, lines, digest \
887
in self.source._data.read_records_iter(records):
888
options = self.source._index.get_options(version_id)
889
parents = self.source._index.get_parents(version_id)
891
for parent in parents:
892
assert self.target.has_version(parent)
894
if self.target.factory.annotated:
895
# FIXME jrydberg: it should be possible to skip
896
# re-annotating components if we know that we are
897
# going to pull all revisions in the same order.
898
new_version_id = version_id
899
new_version_idx = self.target._index.num_versions()
900
if 'fulltext' in options:
901
lines = self.target._reannotate_fulltext(self.source, lines,
902
new_version_id, new_version_idx)
903
elif 'line-delta' in options:
904
lines = self.target._reannotate_line_delta(self.source, lines,
905
new_version_id, new_version_idx)
908
pb.update(self.target.filename, count, len(version_list))
910
pos, size = self.target._data.add_record(version_id, digest, lines)
911
self.target._index.add_version(version_id, options, pos, size, parents)
913
for version in mismatched_versions:
914
n1 = set(self.target.get_parents(version))
915
n2 = set(self.source.get_parents(version))
916
# write a combined record to our history.
917
new_parents = self.target.get_parents(version) + list(n2.difference(n1))
918
current_values = self.target._index._cache[version]
919
self.target._index.add_version(version,
928
InterVersionedFile.register_optimiser(InterKnit)
added
removed
bzrlib/knit.py
