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)
271
def fix_parents(self, version, new_parents):
272
"""Fix the parents list for version.
274
This is done by appending a new version to the index
275
with identical data except for the parents list.
276
the parents list must be a superset of the current
279
current_values = self._index._cache[version]
280
assert set(current_values[4]).difference(set(new_parents)) == set()
281
self._index.add_version(version,
287
def get_graph_with_ghosts(self):
288
"""See VersionedFile.get_graph_with_ghosts()."""
289
graph_items = self._index.get_graph()
290
return dict(graph_items)
294
"""See VersionedFile.get_suffixes()."""
295
return [DATA_SUFFIX, INDEX_SUFFIX]
297
def has_ghost(self, version_id):
298
"""True if there is a ghost reference in the file to version_id."""
300
if self.has_version(version_id):
302
# optimisable if needed by memoising the _ghosts set.
303
items = self._index.get_graph()
304
for node, parents in items:
305
for parent in parents:
306
if parent not in self._index._cache:
307
if parent == version_id:
312
"""See VersionedFile.versions."""
313
return self._index.get_versions()
315
def has_version(self, version_id):
316
"""See VersionedFile.has_version."""
317
return self._index.has_version(version_id)
319
__contains__ = has_version
321
def _merge_annotations(self, content, parents):
322
"""Merge annotations for content. This is done by comparing
323
the annotations based on changed to the text."""
324
for parent_id in parents:
325
merge_content = self._get_content(parent_id)
326
seq = SequenceMatcher(None, merge_content.text(), content.text())
327
for i, j, n in seq.get_matching_blocks():
330
content._lines[j:j+n] = merge_content._lines[i:i+n]
332
def _get_components(self, version_id):
333
"""Return a list of (version_id, method, data) tuples that
334
makes up version specified by version_id of the knit.
336
The components should be applied in the order of the returned
339
The basis knit will be used to the largest extent possible
340
since it is assumed that accesses to it is faster.
342
# needed_revisions holds a list of (method, version_id) of
343
# versions that is needed to be fetched to construct the final
344
# version of the file.
346
# basis_revisions is a list of versions that needs to be
347
# fetched but exists in the basis knit.
349
basis = self.basis_knit
356
if basis and basis._index.has_version(cursor):
358
basis_versions.append(cursor)
359
method = picked_knit._index.get_method(cursor)
360
needed_versions.append((method, cursor))
361
if method == 'fulltext':
363
cursor = picked_knit.get_parents(cursor)[0]
368
for comp_id in basis_versions:
369
data_pos, data_size = basis._index.get_data_position(comp_id)
370
records.append((piece_id, data_pos, data_size))
371
components.update(basis._data.read_records(records))
374
for comp_id in [vid for method, vid in needed_versions
375
if vid not in basis_versions]:
376
data_pos, data_size = self._index.get_position(comp_id)
377
records.append((comp_id, data_pos, data_size))
378
components.update(self._data.read_records(records))
380
# get_data_records returns a mapping with the version id as
381
# index and the value as data. The order the components need
382
# to be applied is held by needed_versions (reversed).
384
for method, comp_id in reversed(needed_versions):
385
out.append((comp_id, method, components[comp_id]))
389
def _get_content(self, version_id):
390
"""Returns a content object that makes up the specified
392
if not self.has_version(version_id):
393
raise RevisionNotPresent(version_id, self.filename)
395
if self.basis_knit and version_id in self.basis_knit:
396
return self.basis_knit._get_content(version_id)
399
components = self._get_components(version_id)
400
for component_id, method, (data, digest) in components:
401
version_idx = self._index.lookup(component_id)
402
if method == 'fulltext':
403
assert content is None
404
content = self.factory.parse_fulltext(data, version_idx)
405
elif method == 'line-delta':
406
delta = self.factory.parse_line_delta(data, version_idx)
407
content.apply_delta(delta)
409
if 'no-eol' in self._index.get_options(version_id):
410
line = content._lines[-1][1].rstrip('\n')
411
content._lines[-1] = (content._lines[-1][0], line)
413
if sha_strings(content.text()) != digest:
414
raise KnitCorrupt(self.filename, 'sha-1 does not match')
418
def _check_versions_present(self, version_ids):
419
"""Check that all specified versions are present."""
420
version_ids = set(version_ids)
421
for r in list(version_ids):
422
if self._index.has_version(r):
423
version_ids.remove(r)
425
raise RevisionNotPresent(list(version_ids)[0], self.filename)
427
def add_lines_with_ghosts(self, version_id, parents, lines):
428
"""See VersionedFile.add_lines_with_ghosts()."""
429
self._check_add(version_id, lines)
430
return self._add(version_id, lines[:], parents, self.delta)
432
def add_lines(self, version_id, parents, lines):
433
"""See VersionedFile.add_lines."""
434
self._check_add(version_id, lines)
435
self._check_versions_present(parents)
436
return self._add(version_id, lines[:], parents, self.delta)
438
def _check_add(self, version_id, lines):
439
"""check that version_id and lines are safe to add."""
440
assert self.writable, "knit is not opened for write"
441
### FIXME escape. RBC 20060228
442
if contains_whitespace(version_id):
443
raise InvalidRevisionId(version_id)
444
if self.has_version(version_id):
445
raise RevisionAlreadyPresent(version_id, self.filename)
447
if False or __debug__:
449
assert '\n' not in l[:-1]
451
def _add(self, version_id, lines, parents, delta):
452
"""Add a set of lines on top of version specified by parents.
454
If delta is true, compress the text as a line-delta against
457
Any versions not present will be converted into ghosts.
459
ghostless_parents = []
461
for parent in parents:
462
if not self.has_version(parent):
463
ghosts.append(parent)
465
ghostless_parents.append(parent)
467
if delta and not len(ghostless_parents):
470
digest = sha_strings(lines)
473
if lines[-1][-1] != '\n':
474
options.append('no-eol')
475
lines[-1] = lines[-1] + '\n'
477
lines = self.factory.make(lines, len(self._index))
478
if self.factory.annotated and len(ghostless_parents) > 0:
479
# Merge annotations from parent texts if so is needed.
480
self._merge_annotations(lines, ghostless_parents)
482
if len(ghostless_parents) and delta:
483
# To speed the extract of texts the delta chain is limited
484
# to a fixed number of deltas. This should minimize both
485
# I/O and the time spend applying deltas.
487
delta_parents = ghostless_parents
489
parent = delta_parents[0]
490
method = self._index.get_method(parent)
491
if method == 'fulltext':
493
delta_parents = self._index.get_parents(parent)
495
if method == 'line-delta':
499
options.append('line-delta')
500
content = self._get_content(ghostless_parents[0])
501
delta_hunks = content.line_delta(lines)
502
store_lines = self.factory.lower_line_delta(delta_hunks)
504
options.append('fulltext')
505
store_lines = self.factory.lower_fulltext(lines)
507
where, size = self._data.add_record(version_id, digest, store_lines)
508
self._index.add_version(version_id, options, where, size, parents)
510
def check(self, progress_bar=None):
511
"""See VersionedFile.check()."""
513
def clone_text(self, new_version_id, old_version_id, parents):
514
"""See VersionedFile.clone_text()."""
515
# FIXME RBC 20060228 make fast by only inserting an index with null delta.
516
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
518
def get_lines(self, version_id):
519
"""See VersionedFile.get_lines()."""
520
return self._get_content(version_id).text()
522
def iter_lines_added_or_present_in_versions(self, version_ids=None):
523
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
524
if version_ids is None:
525
version_ids = self.versions()
526
# we dont care about inclusions, the caller cares.
527
# but we need to setup a list of records to visit.
528
# we need version_id, position, length
529
version_id_records = []
530
for version_id in version_ids:
531
if not self.has_version(version_id):
532
raise RevisionNotPresent(version_id, self.filename)
533
data_pos, length = self._index.get_position(version_id)
534
version_id_records.append((version_id, data_pos, length))
535
for version_id, data, sha_value in \
536
self._data.read_records_iter(version_id_records):
537
method = self._index.get_method(version_id)
538
version_idx = self._index.lookup(version_id)
539
assert method in ('fulltext', 'line-delta')
540
if method == 'fulltext':
541
content = self.factory.parse_fulltext(data, version_idx)
542
for line in content.text():
545
delta = self.factory.parse_line_delta(data, version_idx)
546
for start, end, count, lines in delta:
547
for origin, line in lines:
550
def num_versions(self):
551
"""See VersionedFile.num_versions()."""
552
return self._index.num_versions()
554
__len__ = num_versions
556
def annotate_iter(self, version_id):
557
"""See VersionedFile.annotate_iter."""
558
content = self._get_content(version_id)
559
for origin, text in content.annotate_iter():
560
yield self._index.idx_to_name(origin), text
562
def get_parents(self, version_id):
563
"""See VersionedFile.get_parents."""
564
self._check_versions_present([version_id])
565
return list(self._index.get_parents(version_id))
567
def get_parents_with_ghosts(self, version_id):
568
"""See VersionedFile.get_parents."""
569
self._check_versions_present([version_id])
570
return list(self._index.get_parents_with_ghosts(version_id))
572
def get_ancestry(self, versions):
573
"""See VersionedFile.get_ancestry."""
574
if isinstance(versions, basestring):
575
versions = [versions]
578
self._check_versions_present(versions)
579
return self._index.get_ancestry(versions)
581
def get_ancestry_with_ghosts(self, versions):
582
"""See VersionedFile.get_ancestry_with_ghosts."""
583
if isinstance(versions, basestring):
584
versions = [versions]
587
self._check_versions_present(versions)
588
return self._index.get_ancestry_with_ghosts(versions)
590
def _reannotate_line_delta(self, other, lines, new_version_id,
592
"""Re-annotate line-delta and return new delta."""
594
for start, end, count, contents \
595
in self.factory.parse_line_delta_iter(lines):
597
for origin, line in contents:
598
old_version_id = other._index.idx_to_name(origin)
599
if old_version_id == new_version_id:
600
idx = new_version_idx
602
idx = self._index.lookup(old_version_id)
603
new_lines.append((idx, line))
604
new_delta.append((start, end, count, new_lines))
606
return self.factory.lower_line_delta(new_delta)
608
def _reannotate_fulltext(self, other, lines, new_version_id,
610
"""Re-annotate fulltext and return new version."""
611
content = self.factory.parse_fulltext(lines, new_version_idx)
613
for origin, line in content.annotate_iter():
614
old_version_id = other._index.idx_to_name(origin)
615
if old_version_id == new_version_id:
616
idx = new_version_idx
618
idx = self._index.lookup(old_version_id)
619
new_lines.append((idx, line))
621
return self.factory.lower_fulltext(KnitContent(new_lines))
623
#@deprecated_method(zero_eight)
624
def walk(self, version_ids):
625
"""See VersionedFile.walk."""
626
# We take the short path here, and extract all relevant texts
627
# and put them in a weave and let that do all the work. Far
628
# from optimal, but is much simpler.
629
# FIXME RB 20060228 this really is inefficient!
630
from bzrlib.weave import Weave
632
w = Weave(self.filename)
633
ancestry = self.get_ancestry(version_ids)
634
sorted_graph = topo_sort(self._index.get_graph())
635
version_list = [vid for vid in sorted_graph if vid in ancestry]
637
for version_id in version_list:
638
lines = self.get_lines(version_id)
639
w.add_lines(version_id, self.get_parents(version_id), lines)
641
for lineno, insert_id, dset, line in w.walk(version_ids):
642
yield lineno, insert_id, dset, line
645
class _KnitComponentFile(object):
646
"""One of the files used to implement a knit database"""
648
def __init__(self, transport, filename, mode):
649
self._transport = transport
650
self._filename = filename
653
def write_header(self):
654
old_len = self._transport.append(self._filename, StringIO(self.HEADER))
656
raise KnitCorrupt(self._filename, 'misaligned after writing header')
658
def check_header(self, fp):
659
line = fp.read(len(self.HEADER))
660
if line != self.HEADER:
661
raise KnitHeaderError(badline=line)
664
"""Commit is a nop."""
667
return '%s(%s)' % (self.__class__.__name__, self._filename)
670
class _KnitIndex(_KnitComponentFile):
671
"""Manages knit index file.
673
The index is already kept in memory and read on startup, to enable
674
fast lookups of revision information. The cursor of the index
675
file is always pointing to the end, making it easy to append
678
_cache is a cache for fast mapping from version id to a Index
681
_history is a cache for fast mapping from indexes to version ids.
683
The index data format is dictionary compressed when it comes to
684
parent references; a index entry may only have parents that with a
685
lover index number. As a result, the index is topological sorted.
687
Duplicate entries may be written to the index for a single version id
688
if this is done then the latter one completely replaces the former:
689
this allows updates to correct version and parent information.
690
Note that the two entries may share the delta, and that successive
691
annotations and references MUST point to the first entry.
694
HEADER = "# bzr knit index 7\n"
696
def _cache_version(self, version_id, options, pos, size, parents):
697
val = (version_id, options, pos, size, parents)
698
self._cache[version_id] = val
699
if not version_id in self._history:
700
self._history.append(version_id)
702
def _iter_index(self, fp):
704
for l in lines.splitlines(False):
707
def __init__(self, transport, filename, mode, create=False):
708
_KnitComponentFile.__init__(self, transport, filename, mode)
710
# position in _history is the 'official' index for a revision
711
# but the values may have come from a newer entry.
712
# so - wc -l of a knit index is != the number of uniqe names
716
fp = self._transport.get(self._filename)
717
self.check_header(fp)
718
for rec in self._iter_index(fp):
719
parents = self._parse_parents(rec[4:])
720
self._cache_version(rec[0], rec[1].split(','), int(rec[2]), int(rec[3]),
722
except NoSuchFile, e:
723
if mode != 'w' or not create:
727
def _parse_parents(self, compressed_parents):
728
"""convert a list of string parent values into version ids.
730
ints are looked up in the index.
731
.FOO values are ghosts and converted in to FOO.
734
for value in compressed_parents:
735
if value.startswith('.'):
736
result.append(value[1:])
738
assert isinstance(value, str)
739
result.append(self._history[int(value)])
744
for version_id, index in self._cache.iteritems():
745
graph.append((version_id, index[4]))
748
def get_ancestry(self, versions):
749
"""See VersionedFile.get_ancestry."""
750
# get a graph of all the mentioned versions:
752
pending = set(versions)
754
version = pending.pop()
755
parents = self._cache[version][4]
758
parents = [parent for parent in parents if parent in self._cache]
759
for parent in parents:
760
# if not completed and not a ghost
761
if parent not in graph:
763
graph[version] = parents
764
return topo_sort(graph.items())
766
def get_ancestry_with_ghosts(self, versions):
767
"""See VersionedFile.get_ancestry_with_ghosts."""
768
# get a graph of all the mentioned versions:
770
pending = set(versions)
772
version = pending.pop()
774
parents = self._cache[version][4]
781
for parent in parents:
782
if parent not in graph:
784
graph[version] = parents
785
return topo_sort(graph.items())
787
def num_versions(self):
788
return len(self._history)
790
__len__ = num_versions
792
def get_versions(self):
795
def idx_to_name(self, idx):
796
return self._history[idx]
798
def lookup(self, version_id):
799
assert version_id in self._cache
800
return self._history.index(version_id)
802
def _version_list_to_index(self, versions):
804
for version in versions:
805
if version in self._cache:
806
result_list.append(str(self._history.index(version)))
808
result_list.append('.' + version.encode('utf-8'))
809
return ' '.join(result_list)
811
def add_version(self, version_id, options, pos, size, parents):
812
"""Add a version record to the index."""
813
self._cache_version(version_id, options, pos, size, parents)
815
content = "%s %s %s %s %s\n" % (version_id,
819
self._version_list_to_index(parents))
820
self._transport.append(self._filename, StringIO(content))
822
def has_version(self, version_id):
823
"""True if the version is in the index."""
824
return self._cache.has_key(version_id)
826
def get_position(self, version_id):
827
"""Return data position and size of specified version."""
828
return (self._cache[version_id][2], \
829
self._cache[version_id][3])
831
def get_method(self, version_id):
832
"""Return compression method of specified version."""
833
options = self._cache[version_id][1]
834
if 'fulltext' in options:
837
assert 'line-delta' in options
840
def get_options(self, version_id):
841
return self._cache[version_id][1]
843
def get_parents(self, version_id):
844
"""Return parents of specified version ignoring ghosts."""
845
return [parent for parent in self._cache[version_id][4]
846
if parent in self._cache]
848
def get_parents_with_ghosts(self, version_id):
849
"""Return parents of specified version wth ghosts."""
850
return self._cache[version_id][4]
852
def check_versions_present(self, version_ids):
853
"""Check that all specified versions are present."""
854
version_ids = set(version_ids)
855
for version_id in list(version_ids):
856
if version_id in self._cache:
857
version_ids.remove(version_id)
859
raise RevisionNotPresent(list(version_ids)[0], self.filename)
862
class _KnitData(_KnitComponentFile):
863
"""Contents of the knit data file"""
865
HEADER = "# bzr knit data 7\n"
867
def __init__(self, transport, filename, mode, create=False):
868
_KnitComponentFile.__init__(self, transport, filename, mode)
870
self._checked = False
872
self._transport.put(self._filename, StringIO(''))
874
def _open_file(self):
875
if self._file is None:
877
self._file = self._transport.get(self._filename)
882
def add_record(self, version_id, digest, lines):
883
"""Write new text record to disk. Returns the position in the
884
file where it was written."""
886
data_file = GzipFile(None, mode='wb', fileobj=sio)
887
print >>data_file, "version %s %d %s" % (version_id, len(lines), digest)
888
data_file.writelines(lines)
889
print >>data_file, "end %s\n" % version_id
892
content = sio.getvalue()
893
start_pos = self._transport.append(self._filename, StringIO(content))
894
return start_pos, len(content)
896
def _parse_record(self, version_id, data):
897
df = GzipFile(mode='rb', fileobj=StringIO(data))
898
rec = df.readline().split()
900
raise KnitCorrupt(self._filename, 'unexpected number of records')
901
if rec[1] != version_id:
902
raise KnitCorrupt(self.file.name,
903
'unexpected version, wanted %r' % version_id)
905
record_contents = self._read_record_contents(df, lines)
907
if l != 'end %s\n' % version_id:
908
raise KnitCorrupt(self._filename, 'unexpected version end line %r, wanted %r'
910
return record_contents, rec[3]
912
def _read_record_contents(self, df, record_lines):
913
"""Read and return n lines from datafile."""
915
for i in range(record_lines):
916
r.append(df.readline())
919
def read_records_iter(self, records):
920
"""Read text records from data file and yield result.
922
Each passed record is a tuple of (version_id, pos, len) and
923
will be read in the given order. Yields (version_id,
927
class ContinuousRange:
928
def __init__(self, rec_id, pos, size):
930
self.end_pos = pos + size
931
self.versions = [(rec_id, pos, size)]
933
def add(self, rec_id, pos, size):
934
if self.end_pos != pos:
936
self.end_pos = pos + size
937
self.versions.append((rec_id, pos, size))
941
for rec_id, pos, size in self.versions:
942
yield rec_id, fp.read(size)
944
# We take it that the transport optimizes the fetching as good
945
# as possible (ie, reads continous ranges.)
946
response = self._transport.readv(self._filename,
947
[(pos, size) for version_id, pos, size in records])
949
for (record_id, pos, size), (pos, data) in zip(records, response):
950
content, digest = self._parse_record(record_id, data)
951
yield record_id, content, digest
953
def read_records(self, records):
954
"""Read records into a dictionary."""
956
for record_id, content, digest in self.read_records_iter(records):
957
components[record_id] = (content, digest)
961
class InterKnit(InterVersionedFile):
962
"""Optimised code paths for knit to knit operations."""
964
_matching_file_factory = KnitVersionedFile
967
def is_compatible(source, target):
968
"""Be compatible with knits. """
970
return (isinstance(source, KnitVersionedFile) and
971
isinstance(target, KnitVersionedFile))
972
except AttributeError:
975
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
976
"""See InterVersionedFile.join."""
977
assert isinstance(self.source, KnitVersionedFile)
978
assert isinstance(self.target, KnitVersionedFile)
980
if version_ids is None:
981
version_ids = self.source.versions()
983
if not ignore_missing:
984
self.source._check_versions_present(version_ids)
986
version_ids = set(self.source.versions()).intersection(
993
from bzrlib.progress import DummyProgress
996
version_ids = list(version_ids)
997
if None in version_ids:
998
version_ids.remove(None)
1000
self.source_ancestry = set(self.source.get_ancestry(version_ids))
1001
this_versions = set(self.target._index.get_versions())
1002
needed_versions = self.source_ancestry - this_versions
1003
cross_check_versions = self.source_ancestry.intersection(this_versions)
1004
mismatched_versions = set()
1005
for version in cross_check_versions:
1006
# scan to include needed parents.
1007
n1 = set(self.target.get_parents_with_ghosts(version))
1008
n2 = set(self.source.get_parents_with_ghosts(version))
1010
# FIXME TEST this check for cycles being introduced works
1011
# the logic is we have a cycle if in our graph we are an
1012
# ancestor of any of the n2 revisions.
1018
parent_ancestors = self.source.get_ancestry(parent)
1019
if version in parent_ancestors:
1020
raise errors.GraphCycleError([parent, version])
1021
# ensure this parent will be available later.
1022
new_parents = n2.difference(n1)
1023
needed_versions.update(new_parents.difference(this_versions))
1024
mismatched_versions.add(version)
1026
if not needed_versions and not cross_check_versions:
1028
full_list = topo_sort(self.source.get_graph())
1030
version_list = [i for i in full_list if (not self.target.has_version(i)
1031
and i in needed_versions)]
1034
for version_id in version_list:
1035
data_pos, data_size = self.source._index.get_position(version_id)
1036
records.append((version_id, data_pos, data_size))
1039
for version_id, lines, digest \
1040
in self.source._data.read_records_iter(records):
1041
options = self.source._index.get_options(version_id)
1042
parents = self.source._index.get_parents_with_ghosts(version_id)
1044
for parent in parents:
1045
# if source has the parent, we must hav grabbed it first.
1046
assert (self.target.has_version(parent) or not
1047
self.source.has_version(parent))
1049
if self.target.factory.annotated:
1050
# FIXME jrydberg: it should be possible to skip
1051
# re-annotating components if we know that we are
1052
# going to pull all revisions in the same order.
1053
new_version_id = version_id
1054
new_version_idx = self.target._index.num_versions()
1055
if 'fulltext' in options:
1056
lines = self.target._reannotate_fulltext(self.source, lines,
1057
new_version_id, new_version_idx)
1058
elif 'line-delta' in options:
1059
lines = self.target._reannotate_line_delta(self.source, lines,
1060
new_version_id, new_version_idx)
1063
pb.update("Joining knit", count, len(version_list))
1065
pos, size = self.target._data.add_record(version_id, digest, lines)
1066
self.target._index.add_version(version_id, options, pos, size, parents)
1068
for version in mismatched_versions:
1069
n1 = set(self.target.get_parents_with_ghosts(version))
1070
n2 = set(self.source.get_parents_with_ghosts(version))
1071
# write a combined record to our history preserving the current
1072
# parents as first in the list
1073
new_parents = self.target.get_parents_with_ghosts(version) + list(n2.difference(n1))
1074
self.target.fix_parents(version, new_parents)
1079
InterVersionedFile.register_optimiser(InterKnit)
added
removed
bzrlib/knit.py
