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
65
from difflib import SequenceMatcher
66
from gzip import GzipFile
68
from StringIO import StringIO
70
from bzrlib.errors import FileExists, NoSuchFile, KnitError, \
71
InvalidRevisionId, KnitCorrupt, KnitHeaderError, \
72
RevisionNotPresent, RevisionAlreadyPresent
73
from bzrlib.trace import mutter
74
from bzrlib.osutils import contains_whitespace, contains_linebreaks, \
76
from bzrlib.versionedfile import VersionedFile
77
from bzrlib.tsort import topo_sort
80
# TODO: Split out code specific to this format into an associated object.
82
# TODO: Can we put in some kind of value to check that the index and data
83
# files belong together?
85
# TODO: accomodate binaries, perhaps by storing a byte count
87
# TODO: function to check whole file
89
# TODO: atomically append data, then measure backwards from the cursor
90
# position after writing to work out where it was located. we may need to
91
# bypass python file buffering.
94
INDEX_SUFFIX = '.kndx'
97
class KnitContent(object):
98
"""Content of a knit version to which deltas can be applied."""
100
def __init__(self, lines):
103
def annotate_iter(self):
104
"""Yield tuples of (origin, text) for each content line."""
105
for origin, text in self._lines:
109
"""Return a list of (origin, text) tuples."""
110
return list(self.annotate_iter())
112
def apply_delta(self, delta):
113
"""Apply delta to this content."""
115
for start, end, count, lines in delta:
116
self._lines[offset+start:offset+end] = lines
117
offset = offset + (start - end) + count
119
def line_delta_iter(self, new_lines):
120
"""Generate line-based delta from new_lines to this content."""
121
new_texts = [text for origin, text in new_lines._lines]
122
old_texts = [text for origin, text in self._lines]
123
s = difflib.SequenceMatcher(None, old_texts, new_texts)
124
for op in s.get_opcodes():
127
yield (op[1], op[2], op[4]-op[3], new_lines._lines[op[3]:op[4]])
129
def line_delta(self, new_lines):
130
return list(self.line_delta_iter(new_lines))
133
return [text for origin, text in self._lines]
136
class _KnitFactory(object):
137
"""Base factory for creating content objects."""
139
def make(self, lines, version):
140
num_lines = len(lines)
141
return KnitContent(zip([version] * num_lines, lines))
144
class KnitAnnotateFactory(_KnitFactory):
145
"""Factory for creating annotated Content objects."""
149
def parse_fulltext(self, content, version):
152
origin, text = line.split(' ', 1)
153
lines.append((int(origin), text))
154
return KnitContent(lines)
156
def parse_line_delta_iter(self, lines):
158
header = lines.pop(0)
159
start, end, c = [int(n) for n in header.split(',')]
162
origin, text = lines.pop(0).split(' ', 1)
163
contents.append((int(origin), text))
164
yield start, end, c, contents
166
def parse_line_delta(self, lines, version):
167
return list(self.parse_line_delta_iter(lines))
169
def lower_fulltext(self, content):
170
return ['%d %s' % (o, t) for o, t in content._lines]
172
def lower_line_delta(self, delta):
174
for start, end, c, lines in delta:
175
out.append('%d,%d,%d\n' % (start, end, c))
176
for origin, text in lines:
177
out.append('%d %s' % (origin, text))
181
class KnitPlainFactory(_KnitFactory):
182
"""Factory for creating plain Content objects."""
186
def parse_fulltext(self, content, version):
187
return self.make(content, version)
189
def parse_line_delta_iter(self, lines, version):
191
header = lines.pop(0)
192
start, end, c = [int(n) for n in header.split(',')]
193
yield start, end, c, zip([version] * c, lines[:c])
196
def parse_line_delta(self, lines, version):
197
return list(self.parse_line_delta_iter(lines, version))
199
def lower_fulltext(self, content):
200
return content.text()
202
def lower_line_delta(self, delta):
204
for start, end, c, lines in delta:
205
out.append('%d,%d,%d\n' % (start, end, c))
206
out.extend([text for origin, text in lines])
210
def make_empty_knit(transport, relpath):
211
"""Construct a empty knit at the specified location."""
212
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
216
class KnitVersionedFile(VersionedFile):
217
"""Weave-like structure with faster random access.
219
A knit stores a number of texts and a summary of the relationships
220
between them. Texts are identified by a string version-id. Texts
221
are normally stored and retrieved as a series of lines, but can
222
also be passed as single strings.
224
Lines are stored with the trailing newline (if any) included, to
225
avoid special cases for files with no final newline. Lines are
226
composed of 8-bit characters, not unicode. The combination of
227
these approaches should mean any 'binary' file can be safely
228
stored and retrieved.
231
def __init__(self, transport, relpath, mode, factory,
232
basis_knit=None, delta=True):
233
"""Construct a knit at location specified by relpath."""
234
assert mode in ('r', 'w'), "invalid mode specified"
235
assert not basis_knit or isinstance(basis_knit, KnitVersionedFile), \
238
self.transport = transport
239
self.filename = relpath
240
self.basis_knit = basis_knit
241
self.factory = factory
242
self.writable = (mode == 'w')
245
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
247
self._data = _KnitData(transport, relpath + DATA_SUFFIX,
251
"""See VersionedFile.versions."""
252
return self._index.get_versions()
254
def has_version(self, version_id):
255
"""See VersionedFile.has_version."""
256
return self._index.has_version(version_id)
258
__contains__ = has_version
260
def _merge_annotations(self, content, parents):
261
"""Merge annotations for content. This is done by comparing
262
the annotations based on changed to the text."""
263
for parent_id in parents:
264
merge_content = self._get_content(parent_id)
265
seq = SequenceMatcher(None, merge_content.text(), content.text())
266
for i, j, n in seq.get_matching_blocks():
269
content._lines[j:j+n] = merge_content._lines[i:i+n]
271
def _get_components(self, version_id):
272
"""Return a list of (version_id, method, data) tuples that
273
makes up version specified by version_id of the knit.
275
The components should be applied in the order of the returned
278
The basis knit will be used to the largest extent possible
279
since it is assumed that accesses to it is faster.
281
# needed_revisions holds a list of (method, version_id) of
282
# versions that is needed to be fetched to construct the final
283
# version of the file.
285
# basis_revisions is a list of versions that needs to be
286
# fetched but exists in the basis knit.
288
basis = self.basis_knit
295
if basis and basis._index.has_version(cursor):
297
basis_versions.append(cursor)
298
method = picked_knit._index.get_method(cursor)
299
needed_versions.append((method, cursor))
300
if method == 'fulltext':
302
cursor = picked_knit.get_parents(cursor)[0]
307
for comp_id in basis_versions:
308
data_pos, data_size = basis._index.get_data_position(comp_id)
309
records.append((piece_id, data_pos, data_size))
310
components.update(basis._data.read_records(records))
313
for comp_id in [vid for method, vid in needed_versions
314
if vid not in basis_versions]:
315
data_pos, data_size = self._index.get_position(comp_id)
316
records.append((comp_id, data_pos, data_size))
317
components.update(self._data.read_records(records))
319
# get_data_records returns a mapping with the version id as
320
# index and the value as data. The order the components need
321
# to be applied is held by needed_versions (reversed).
323
for method, comp_id in reversed(needed_versions):
324
out.append((comp_id, method, components[comp_id]))
328
def _get_content(self, version_id):
329
"""Returns a content object that makes up the specified
331
if not self.has_version(version_id):
332
raise RevisionNotPresent(version_id, self.filename)
334
if self.basis_knit and version_id in self.basis_knit:
335
return self.basis_knit._get_content(version_id)
338
components = self._get_components(version_id)
339
for component_id, method, (data, digest) in components:
340
version_idx = self._index.lookup(component_id)
341
if method == 'fulltext':
342
assert content is None
343
content = self.factory.parse_fulltext(data, version_idx)
344
elif method == 'line-delta':
345
delta = self.factory.parse_line_delta(data, version_idx)
346
content.apply_delta(delta)
348
if 'no-eol' in self._index.get_options(version_id):
349
line = content._lines[-1][1].rstrip('\n')
350
content._lines[-1] = (content._lines[-1][0], line)
352
if sha_strings(content.text()) != digest:
353
raise KnitCorrupt(self.filename, 'sha-1 does not match')
357
def _check_versions_present(self, version_ids):
358
"""Check that all specified versions are present."""
359
version_ids = set(version_ids)
360
for r in list(version_ids):
361
if self._index.has_version(r):
362
version_ids.remove(r)
364
raise RevisionNotPresent(list(version_ids)[0], self.filename)
366
def add_lines(self, version_id, parents, lines):
367
"""See VersionedFile.add_lines."""
368
assert self.writable, "knit is not opened for write"
369
### FIXME escape. RBC 20060228
370
if contains_whitespace(version_id):
371
raise InvalidRevisionId(version_id)
372
if self.has_version(version_id):
373
raise RevisionAlreadyPresent(version_id, self.filename)
375
if True or __debug__:
377
assert '\n' not in l[:-1]
379
self._check_versions_present(parents)
380
return self._add(version_id, lines[:], parents, self.delta)
382
def _add(self, version_id, lines, parents, delta):
383
"""Add a set of lines on top of version specified by parents.
385
If delta is true, compress the text as a line-delta against
388
if delta and not parents:
391
digest = sha_strings(lines)
394
if lines[-1][-1] != '\n':
395
options.append('no-eol')
396
lines[-1] = lines[-1] + '\n'
398
lines = self.factory.make(lines, len(self._index))
399
if self.factory.annotated and len(parents) > 0:
400
# Merge annotations from parent texts if so is needed.
401
self._merge_annotations(lines, parents)
403
if parents and delta:
404
# To speed the extract of texts the delta chain is limited
405
# to a fixed number of deltas. This should minimize both
406
# I/O and the time spend applying deltas.
408
delta_parents = parents
410
parent = delta_parents[0]
411
method = self._index.get_method(parent)
412
if method == 'fulltext':
414
delta_parents = self._index.get_parents(parent)
416
if method == 'line-delta':
420
options.append('line-delta')
421
content = self._get_content(parents[0])
422
delta_hunks = content.line_delta(lines)
423
store_lines = self.factory.lower_line_delta(delta_hunks)
425
options.append('fulltext')
426
store_lines = self.factory.lower_fulltext(lines)
428
where, size = self._data.add_record(version_id, digest, store_lines)
429
self._index.add_version(version_id, options, where, size, parents)
431
def clone_text(self, new_version_id, old_version_id, parents):
432
"""See VersionedFile.clone_text()."""
433
# FIXME RBC 20060228 make fast by only inserting an index with null delta.
434
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
436
def get_lines(self, version_id):
437
"""See VersionedFile.get_lines()."""
438
return self._get_content(version_id).text()
440
def annotate_iter(self, version_id):
441
"""See VersionedFile.annotate_iter."""
442
content = self._get_content(version_id)
443
for origin, text in content.annotate_iter():
444
yield self._index.idx_to_name(origin), text
446
def get_parents(self, version_id):
447
"""See VersionedFile.get_parents."""
448
self._check_versions_present([version_id])
449
return list(self._index.get_parents(version_id))
451
def get_ancestry(self, versions):
452
"""See VersionedFile.get_ancestry."""
453
if isinstance(versions, basestring):
454
versions = [versions]
457
self._check_versions_present(versions)
458
return self._index.get_ancestry(versions)
460
def _reannotate_line_delta(self, other, lines, new_version_id,
462
"""Re-annotate line-delta and return new delta."""
464
for start, end, count, contents \
465
in self.factory.parse_line_delta_iter(lines):
467
for origin, line in contents:
468
old_version_id = other._index.idx_to_name(origin)
469
if old_version_id == new_version_id:
470
idx = new_version_idx
472
idx = self._index.lookup(old_version_id)
473
new_lines.append((idx, line))
474
new_delta.append((start, end, count, new_lines))
476
return self.factory.lower_line_delta(new_delta)
478
def _reannotate_fulltext(self, other, lines, new_version_id,
480
"""Re-annotate fulltext and return new version."""
481
content = self.factory.parse_fulltext(lines, new_version_idx)
483
for origin, line in content.annotate_iter():
484
old_version_id = other._index.idx_to_name(origin)
485
if old_version_id == new_version_id:
486
idx = new_version_idx
488
idx = self._index.lookup(old_version_id)
489
new_lines.append((idx, line))
491
return self.factory.lower_fulltext(KnitContent(new_lines))
493
def join(self, other, pb=None, msg=None, version_ids=None):
494
"""See VersionedFile.join."""
495
assert isinstance(other, KnitVersionedFile)
497
if version_ids is None:
498
version_ids = other.versions()
503
from bzrlib.progress import DummyProgress
506
version_ids = list(version_ids)
507
if None in version_ids:
508
version_ids.remove(None)
510
other_ancestry = set(other.get_ancestry(version_ids))
511
needed_versions = other_ancestry - set(self._index.get_versions())
512
if not needed_versions:
514
full_list = topo_sort(other._index.get_graph())
516
version_list = [i for i in full_list if (not self.has_version(i)
517
and i in needed_versions)]
520
for version_id in version_list:
521
data_pos, data_size = other._index.get_position(version_id)
522
records.append((version_id, data_pos, data_size))
525
for version_id, lines, digest \
526
in other._data.read_records_iter(records):
527
options = other._index.get_options(version_id)
528
parents = other._index.get_parents(version_id)
530
for parent in parents:
531
assert self.has_version(parent)
533
if self.factory.annotated:
534
# FIXME jrydberg: it should be possible to skip
535
# re-annotating components if we know that we are
536
# going to pull all revisions in the same order.
537
new_version_id = version_id
538
new_version_idx = self._index.num_versions()
539
if 'fulltext' in options:
540
lines = self._reannotate_fulltext(other, lines,
541
new_version_id, new_version_idx)
542
elif 'line-delta' in options:
543
lines = self._reannotate_line_delta(other, lines,
544
new_version_id, new_version_idx)
547
pb.update(self.filename, count, len(version_list))
549
pos, size = self._data.add_record(version_id, digest, lines)
550
self._index.add_version(version_id, options, pos, size, parents)
555
def walk(self, version_ids):
556
"""See VersionedFile.walk."""
557
# We take the short path here, and extract all relevant texts
558
# and put them in a weave and let that do all the work. Far
559
# from optimal, but is much simpler.
560
# FIXME RB 20060228 this really is inefficient!
561
from bzrlib.weave import Weave
563
w = Weave(self.filename)
564
ancestry = self.get_ancestry(version_ids)
565
sorted_graph = topo_sort(self._index.get_graph())
566
version_list = [vid for vid in sorted_graph if vid in ancestry]
568
for version_id in version_list:
569
lines = self.get_lines(version_id)
570
w.add_lines(version_id, self.get_parents(version_id), lines)
572
for lineno, insert_id, dset, line in w.walk(version_ids):
573
yield lineno, insert_id, dset, line
576
class _KnitComponentFile(object):
577
"""One of the files used to implement a knit database"""
579
def __init__(self, transport, filename, mode):
580
self._transport = transport
581
self._filename = filename
584
def write_header(self):
585
old_len = self._transport.append(self._filename, self.HEADER)
587
raise KnitCorrupt(self._filename, 'misaligned after writing header')
589
def check_header(self, fp):
590
line = fp.read(len(self.HEADER))
591
if line != self.HEADER:
592
raise KnitHeaderError(badline=line)
595
"""Commit is a nop."""
598
return '%s(%s)' % (self.__class__.__name__, self._filename)
601
class _KnitIndex(_KnitComponentFile):
602
"""Manages knit index file.
604
The index is already kept in memory and read on startup, to enable
605
fast lookups of revision information. The cursor of the index
606
file is always pointing to the end, making it easy to append
609
_cache is a cache for fast mapping from version id to a Index
612
_history is a cache for fast mapping from indexes to version ids.
614
The index data format is dictionary compressed when it comes to
615
parent references; a index entry may only have parents that with a
616
lover index number. As a result, the index is topological sorted.
619
HEADER = "# bzr knit index 7\n"
621
def _cache_version(self, version_id, options, pos, size, parents):
622
val = (version_id, options, pos, size, parents)
623
self._cache[version_id] = val
624
self._history.append(version_id)
626
def _iter_index(self, fp):
628
for l in lines.splitlines(False):
631
def __init__(self, transport, filename, mode):
632
_KnitComponentFile.__init__(self, transport, filename, mode)
636
fp = self._transport.get(self._filename)
637
self.check_header(fp)
638
for rec in self._iter_index(fp):
639
self._cache_version(rec[0], rec[1].split(','), int(rec[2]), int(rec[3]),
640
[self._history[int(i)] for i in rec[4:]])
641
except NoSuchFile, e:
648
for version_id, index in self._cache.iteritems():
649
graph.append((version_id, index[4]))
652
def get_ancestry(self, versions):
653
"""See VersionedFile.get_ancestry."""
655
for version_id in versions:
656
version_idxs.append(self._history.index(version_id))
658
for v in xrange(max(version_idxs), 0, -1):
659
if self._history[v] in i:
660
# include all its parents
661
i.update(self._cache[self._history[v]][4])
664
def num_versions(self):
665
return len(self._history)
667
__len__ = num_versions
669
def get_versions(self):
672
def idx_to_name(self, idx):
673
return self._history[idx]
675
def lookup(self, version_id):
676
assert version_id in self._cache
677
return self._history.index(version_id)
679
def add_version(self, version_id, options, pos, size, parents):
680
"""Add a version record to the index."""
681
self._cache_version(version_id, options, pos, size, parents)
683
content = "%s %s %s %s %s\n" % (version_id,
687
' '.join([str(self.lookup(vid)) for
689
self._transport.append(self._filename, content)
691
def has_version(self, version_id):
692
"""True if the version is in the index."""
693
return self._cache.has_key(version_id)
695
def get_position(self, version_id):
696
"""Return data position and size of specified version."""
697
return (self._cache[version_id][2], \
698
self._cache[version_id][3])
700
def get_method(self, version_id):
701
"""Return compression method of specified version."""
702
options = self._cache[version_id][1]
703
if 'fulltext' in options:
706
assert 'line-delta' in options
709
def get_options(self, version_id):
710
return self._cache[version_id][1]
712
def get_parents(self, version_id):
713
"""Return parents of specified version."""
714
return self._cache[version_id][4]
716
def check_versions_present(self, version_ids):
717
"""Check that all specified versions are present."""
718
version_ids = set(version_ids)
719
for version_id in list(version_ids):
720
if version_id in self._cache:
721
version_ids.remove(version_id)
723
raise RevisionNotPresent(list(version_ids)[0], self.filename)
726
class _KnitData(_KnitComponentFile):
727
"""Contents of the knit data file"""
729
HEADER = "# bzr knit data 7\n"
731
def __init__(self, transport, filename, mode):
732
_KnitComponentFile.__init__(self, transport, filename, mode)
734
self._checked = False
736
def _open_file(self):
737
if self._file is None:
739
self._file = self._transport.get(self._filename)
744
def add_record(self, version_id, digest, lines):
745
"""Write new text record to disk. Returns the position in the
746
file where it was written."""
748
data_file = GzipFile(None, mode='wb', fileobj=sio)
749
print >>data_file, "version %s %d %s" % (version_id, len(lines), digest)
750
data_file.writelines(lines)
751
print >>data_file, "end %s\n" % version_id
754
content = sio.getvalue()
755
start_pos = self._transport.append(self._filename, content)
756
return start_pos, len(content)
758
def _parse_record(self, version_id, data):
759
df = GzipFile(mode='rb', fileobj=StringIO(data))
760
rec = df.readline().split()
762
raise KnitCorrupt(self._filename, 'unexpected number of records')
763
if rec[1] != version_id:
764
raise KnitCorrupt(self.file.name,
765
'unexpected version, wanted %r' % version_id)
767
record_contents = self._read_record_contents(df, lines)
769
if l != 'end %s\n' % version_id:
770
raise KnitCorrupt(self._filename, 'unexpected version end line %r, wanted %r'
772
return record_contents, rec[3]
774
def _read_record_contents(self, df, record_lines):
775
"""Read and return n lines from datafile."""
777
for i in range(record_lines):
778
r.append(df.readline())
781
def read_records_iter(self, records):
782
"""Read text records from data file and yield result.
784
Each passed record is a tuple of (version_id, pos, len) and
785
will be read in the given order. Yields (version_id,
789
class ContinuousRange:
790
def __init__(self, rec_id, pos, size):
792
self.end_pos = pos + size
793
self.versions = [(rec_id, pos, size)]
795
def add(self, rec_id, pos, size):
796
if self.end_pos != pos:
798
self.end_pos = pos + size
799
self.versions.append((rec_id, pos, size))
803
for rec_id, pos, size in self.versions:
804
yield rec_id, fp.read(size)
806
fp = self._open_file()
808
# Loop through all records and try to collect as large
809
# continuous region as possible to read.
811
record_id, pos, size = records.pop(0)
812
continuous_range = ContinuousRange(record_id, pos, size)
814
record_id, pos, size = records[0]
815
if continuous_range.add(record_id, pos, size):
819
fp.seek(continuous_range.start_pos, 0)
820
for record_id, data in continuous_range.split(fp):
821
content, digest = self._parse_record(record_id, data)
822
yield record_id, content, digest
826
def read_records(self, records):
827
"""Read records into a dictionary."""
829
for record_id, content, digest in self.read_records_iter(records):
830
components[record_id] = (content, digest)
added
removed
bzrlib/knit.py
