248
628
for start, end, c, lines in delta:
249
629
out.append('%d,%d,%d\n' % (start, end, c))
250
out.extend([text for origin, text in lines])
254
def make_empty_knit(transport, relpath):
255
"""Construct a empty knit at the specified location."""
256
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
260
class KnitVersionedFile(VersionedFile):
261
"""Weave-like structure with faster random access.
263
A knit stores a number of texts and a summary of the relationships
264
between them. Texts are identified by a string version-id. Texts
265
are normally stored and retrieved as a series of lines, but can
266
also be passed as single strings.
268
Lines are stored with the trailing newline (if any) included, to
269
avoid special cases for files with no final newline. Lines are
270
composed of 8-bit characters, not unicode. The combination of
271
these approaches should mean any 'binary' file can be safely
272
stored and retrieved.
275
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
276
factory=None, basis_knit=DEPRECATED_PARAMETER, delta=True,
278
"""Construct a knit at location specified by relpath.
280
:param create: If not True, only open an existing knit.
282
if deprecated_passed(basis_knit):
283
warnings.warn("KnitVersionedFile.__(): The basis_knit parameter is"
284
" deprecated as of bzr 0.9.",
285
DeprecationWarning, stacklevel=2)
286
if access_mode is None:
288
super(KnitVersionedFile, self).__init__(access_mode)
289
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
290
self.transport = transport
291
self.filename = relpath
292
self.factory = factory or KnitAnnotateFactory()
293
self.writable = (access_mode == 'w')
296
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
297
access_mode, create=create, file_mode=file_mode)
298
self._data = _KnitData(transport, relpath + DATA_SUFFIX,
299
access_mode, create=create and not len(self), file_mode=file_mode)
302
return '%s(%s)' % (self.__class__.__name__,
303
self.transport.abspath(self.filename))
305
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
306
"""See VersionedFile._add_delta()."""
307
self._check_add(version_id, []) # should we check the lines ?
308
self._check_versions_present(parents)
312
for parent in parents:
313
if not self.has_version(parent):
314
ghosts.append(parent)
316
present_parents.append(parent)
318
if delta_parent is None:
319
# reconstitute as full text.
320
assert len(delta) == 1 or len(delta) == 0
322
assert delta[0][0] == 0
323
assert delta[0][1] == 0, delta[0][1]
324
return super(KnitVersionedFile, self)._add_delta(version_id,
335
options.append('no-eol')
337
if delta_parent is not None:
338
# determine the current delta chain length.
339
# To speed the extract of texts the delta chain is limited
340
# to a fixed number of deltas. This should minimize both
341
# I/O and the time spend applying deltas.
343
delta_parents = [delta_parent]
345
parent = delta_parents[0]
346
method = self._index.get_method(parent)
347
if method == 'fulltext':
349
delta_parents = self._index.get_parents(parent)
351
if method == 'line-delta':
352
# did not find a fulltext in the delta limit.
353
# just do a normal insertion.
354
return super(KnitVersionedFile, self)._add_delta(version_id,
361
options.append('line-delta')
362
store_lines = self.factory.lower_line_delta(delta)
364
where, size = self._data.add_record(version_id, digest, store_lines)
365
self._index.add_version(version_id, options, where, size, parents)
367
def _add_raw_records(self, records, data):
368
"""Add all the records 'records' with data pre-joined in 'data'.
370
:param records: A list of tuples(version_id, options, parents, size).
371
:param data: The data for the records. When it is written, the records
372
are adjusted to have pos pointing into data by the sum of
373
the preceding records sizes.
376
pos = self._data.add_raw_record(data)
378
for (version_id, options, parents, size) in records:
379
index_entries.append((version_id, options, pos, size, parents))
381
self._index.add_versions(index_entries)
383
def clear_cache(self):
384
"""Clear the data cache only."""
385
self._data.clear_cache()
387
def copy_to(self, name, transport):
388
"""See VersionedFile.copy_to()."""
389
# copy the current index to a temp index to avoid racing with local
391
transport.put(name + INDEX_SUFFIX + '.tmp', self.transport.get(self._index._filename),)
393
f = self._data._open_file()
395
transport.put(name + DATA_SUFFIX, f)
398
# move the copied index into place
399
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
401
def create_empty(self, name, transport, mode=None):
402
return KnitVersionedFile(name, transport, factory=self.factory, delta=self.delta, create=True)
404
def _fix_parents(self, version, new_parents):
405
"""Fix the parents list for version.
407
This is done by appending a new version to the index
408
with identical data except for the parents list.
409
the parents list must be a superset of the current
412
current_values = self._index._cache[version]
413
assert set(current_values[4]).difference(set(new_parents)) == set()
414
self._index.add_version(version,
420
def get_delta(self, version_id):
421
"""Get a delta for constructing version from some other version."""
422
if not self.has_version(version_id):
423
raise RevisionNotPresent(version_id, self.filename)
425
parents = self.get_parents(version_id)
430
data_pos, data_size = self._index.get_position(version_id)
431
data, sha1 = self._data.read_records(((version_id, data_pos, data_size),))[version_id]
432
version_idx = self._index.lookup(version_id)
433
noeol = 'no-eol' in self._index.get_options(version_id)
434
if 'fulltext' == self._index.get_method(version_id):
435
new_content = self.factory.parse_fulltext(data, version_idx)
436
if parent is not None:
437
reference_content = self._get_content(parent)
438
old_texts = reference_content.text()
441
new_texts = new_content.text()
442
delta_seq = KnitSequenceMatcher(None, old_texts, new_texts)
443
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
445
delta = self.factory.parse_line_delta(data, version_idx)
446
return parent, sha1, noeol, delta
448
def get_graph_with_ghosts(self):
449
"""See VersionedFile.get_graph_with_ghosts()."""
450
graph_items = self._index.get_graph()
451
return dict(graph_items)
453
def get_sha1(self, version_id):
454
"""See VersionedFile.get_sha1()."""
455
record_map = self._get_record_map([version_id])
456
method, content, digest, next = record_map[version_id]
461
"""See VersionedFile.get_suffixes()."""
462
return [DATA_SUFFIX, INDEX_SUFFIX]
464
def has_ghost(self, version_id):
465
"""True if there is a ghost reference in the file to version_id."""
467
if self.has_version(version_id):
469
# optimisable if needed by memoising the _ghosts set.
470
items = self._index.get_graph()
471
for node, parents in items:
472
for parent in parents:
473
if parent not in self._index._cache:
474
if parent == version_id:
479
"""See VersionedFile.versions."""
480
return self._index.get_versions()
482
def has_version(self, version_id):
483
"""See VersionedFile.has_version."""
484
return self._index.has_version(version_id)
486
__contains__ = has_version
488
def _merge_annotations(self, content, parents, parent_texts={},
489
delta=None, annotated=None):
490
"""Merge annotations for content. This is done by comparing
491
the annotations based on changed to the text.
633
def annotate(self, knit, key):
634
annotator = _KnitAnnotator(knit)
635
return annotator.annotate(key)
639
def make_file_factory(annotated, mapper):
640
"""Create a factory for creating a file based KnitVersionedFiles.
642
This is only functional enough to run interface tests, it doesn't try to
643
provide a full pack environment.
645
:param annotated: knit annotations are wanted.
646
:param mapper: The mapper from keys to paths.
648
def factory(transport):
649
index = _KndxIndex(transport, mapper, lambda:None, lambda:True, lambda:True)
650
access = _KnitKeyAccess(transport, mapper)
651
return KnitVersionedFiles(index, access, annotated=annotated)
655
def make_pack_factory(graph, delta, keylength):
656
"""Create a factory for creating a pack based VersionedFiles.
658
This is only functional enough to run interface tests, it doesn't try to
659
provide a full pack environment.
661
:param graph: Store a graph.
662
:param delta: Delta compress contents.
663
:param keylength: How long should keys be.
665
def factory(transport):
666
parents = graph or delta
672
max_delta_chain = 200
675
graph_index = _mod_index.InMemoryGraphIndex(reference_lists=ref_length,
676
key_elements=keylength)
677
stream = transport.open_write_stream('newpack')
678
writer = pack.ContainerWriter(stream.write)
680
index = _KnitGraphIndex(graph_index, lambda:True, parents=parents,
681
deltas=delta, add_callback=graph_index.add_nodes)
682
access = _DirectPackAccess({})
683
access.set_writer(writer, graph_index, (transport, 'newpack'))
684
result = KnitVersionedFiles(index, access,
685
max_delta_chain=max_delta_chain)
686
result.stream = stream
687
result.writer = writer
692
def cleanup_pack_knit(versioned_files):
693
versioned_files.stream.close()
694
versioned_files.writer.end()
697
class KnitVersionedFiles(VersionedFiles):
698
"""Storage for many versioned files using knit compression.
700
Backend storage is managed by indices and data objects.
702
:ivar _index: A _KnitGraphIndex or similar that can describe the
703
parents, graph, compression and data location of entries in this
704
KnitVersionedFiles. Note that this is only the index for
705
*this* vfs; if there are fallbacks they must be queried separately.
708
def __init__(self, index, data_access, max_delta_chain=200,
710
"""Create a KnitVersionedFiles with index and data_access.
712
:param index: The index for the knit data.
713
:param data_access: The access object to store and retrieve knit
715
:param max_delta_chain: The maximum number of deltas to permit during
716
insertion. Set to 0 to prohibit the use of deltas.
717
:param annotated: Set to True to cause annotations to be calculated and
718
stored during insertion.
721
self._access = data_access
722
self._max_delta_chain = max_delta_chain
495
for parent_id in parents:
496
merge_content = self._get_content(parent_id, parent_texts)
497
seq = KnitSequenceMatcher(None, merge_content.text(), content.text())
498
if delta_seq is None:
499
# setup a delta seq to reuse.
501
for i, j, n in seq.get_matching_blocks():
504
# this appears to copy (origin, text) pairs across to the new
505
# content for any line that matches the last-checked parent.
506
# FIXME: save the sequence control data for delta compression
507
# against the most relevant parent rather than rediffing.
508
content._lines[j:j+n] = merge_content._lines[i:i+n]
511
reference_content = self._get_content(parents[0], parent_texts)
512
new_texts = content.text()
513
old_texts = reference_content.text()
514
delta_seq = KnitSequenceMatcher(None, old_texts, new_texts)
515
return self._make_line_delta(delta_seq, content)
517
def _make_line_delta(self, delta_seq, new_content):
518
"""Generate a line delta from delta_seq and new_content."""
520
for op in delta_seq.get_opcodes():
523
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
526
def _get_components_positions(self, version_ids):
527
"""Produce a map of position data for the components of versions.
529
This data is intended to be used for retrieving the knit records.
531
A dict of version_id to (method, data_pos, data_size, next) is
533
method is the way referenced data should be applied.
534
data_pos is the position of the data in the knit.
535
data_size is the size of the data in the knit.
536
next is the build-parent of the version, or None for fulltexts.
724
self._factory = KnitAnnotateFactory()
726
self._factory = KnitPlainFactory()
727
self._fallback_vfs = []
729
def add_fallback_versioned_files(self, a_versioned_files):
730
"""Add a source of texts for texts not present in this knit.
732
:param a_versioned_files: A VersionedFiles object.
539
for version_id in version_ids:
542
while cursor is not None and cursor not in component_data:
543
method = self._index.get_method(cursor)
544
if method == 'fulltext':
547
next = self.get_parents(cursor)[0]
548
data_pos, data_size = self._index.get_position(cursor)
549
component_data[cursor] = (method, data_pos, data_size, next)
551
return component_data
553
def _get_content(self, version_id, parent_texts={}):
554
"""Returns a content object that makes up the specified
556
if not self.has_version(version_id):
557
raise RevisionNotPresent(version_id, self.filename)
559
cached_version = parent_texts.get(version_id, None)
560
if cached_version is not None:
561
return cached_version
563
text_map, contents_map = self._get_content_maps([version_id])
564
return contents_map[version_id]
566
def _check_versions_present(self, version_ids):
567
"""Check that all specified versions are present."""
568
version_ids = set(version_ids)
569
for r in list(version_ids):
570
if self._index.has_version(r):
571
version_ids.remove(r)
573
raise RevisionNotPresent(list(version_ids)[0], self.filename)
575
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
576
"""See VersionedFile.add_lines_with_ghosts()."""
577
self._check_add(version_id, lines)
578
return self._add(version_id, lines[:], parents, self.delta, parent_texts)
580
def _add_lines(self, version_id, parents, lines, parent_texts):
581
"""See VersionedFile.add_lines."""
582
self._check_add(version_id, lines)
583
self._check_versions_present(parents)
584
return self._add(version_id, lines[:], parents, self.delta, parent_texts)
586
def _check_add(self, version_id, lines):
587
"""check that version_id and lines are safe to add."""
588
assert self.writable, "knit is not opened for write"
589
### FIXME escape. RBC 20060228
590
if contains_whitespace(version_id):
591
raise InvalidRevisionId(version_id, self.filename)
592
if self.has_version(version_id):
593
raise RevisionAlreadyPresent(version_id, self.filename)
594
self._check_lines_not_unicode(lines)
595
self._check_lines_are_lines(lines)
597
def _add(self, version_id, lines, parents, delta, parent_texts):
734
self._fallback_vfs.append(a_versioned_files)
736
def add_lines(self, key, parents, lines, parent_texts=None,
737
left_matching_blocks=None, nostore_sha=None, random_id=False,
739
"""See VersionedFiles.add_lines()."""
740
self._index._check_write_ok()
741
self._check_add(key, lines, random_id, check_content)
743
# The caller might pass None if there is no graph data, but kndx
744
# indexes can't directly store that, so we give them
745
# an empty tuple instead.
747
return self._add(key, lines, parents,
748
parent_texts, left_matching_blocks, nostore_sha, random_id)
750
def _add(self, key, lines, parents, parent_texts,
751
left_matching_blocks, nostore_sha, random_id):
598
752
"""Add a set of lines on top of version specified by parents.
600
If delta is true, compress the text as a line-delta against
603
754
Any versions not present will be converted into ghosts.
605
# 461 0 6546.0390 43.9100 bzrlib.knit:489(_add)
606
# +400 0 889.4890 418.9790 +bzrlib.knit:192(lower_fulltext)
607
# +461 0 1364.8070 108.8030 +bzrlib.knit:996(add_record)
608
# +461 0 193.3940 41.5720 +bzrlib.knit:898(add_version)
609
# +461 0 134.0590 18.3810 +bzrlib.osutils:361(sha_strings)
610
# +461 0 36.3420 15.4540 +bzrlib.knit:146(make)
611
# +1383 0 8.0370 8.0370 +<len>
612
# +61 0 13.5770 7.9190 +bzrlib.knit:199(lower_line_delta)
613
# +61 0 963.3470 7.8740 +bzrlib.knit:427(_get_content)
614
# +61 0 973.9950 5.2950 +bzrlib.knit:136(line_delta)
615
# +61 0 1918.1800 5.2640 +bzrlib.knit:359(_merge_annotations)
756
# first thing, if the content is something we don't need to store, find
758
line_bytes = ''.join(lines)
759
digest = sha_string(line_bytes)
760
if nostore_sha == digest:
761
raise errors.ExistingContent
617
763
present_parents = []
619
764
if parent_texts is None:
620
765
parent_texts = {}
766
# Do a single query to ascertain parent presence.
767
present_parent_map = self.get_parent_map(parents)
621
768
for parent in parents:
622
if not self.has_version(parent):
623
ghosts.append(parent)
769
if parent in present_parent_map:
625
770
present_parents.append(parent)
627
if delta and not len(present_parents):
772
# Currently we can only compress against the left most present parent.
773
if (len(present_parents) == 0 or
774
present_parents[0] != parents[0]):
777
# To speed the extract of texts the delta chain is limited
778
# to a fixed number of deltas. This should minimize both
779
# I/O and the time spend applying deltas.
780
delta = self._check_should_delta(present_parents[0])
630
digest = sha_strings(lines)
782
text_length = len(line_bytes)
633
785
if lines[-1][-1] != '\n':
786
# copy the contents of lines.
634
788
options.append('no-eol')
635
789
lines[-1] = lines[-1] + '\n'
637
if len(present_parents) and delta:
638
# To speed the extract of texts the delta chain is limited
639
# to a fixed number of deltas. This should minimize both
640
# I/O and the time spend applying deltas.
642
delta_parents = present_parents
644
parent = delta_parents[0]
645
method = self._index.get_method(parent)
646
if method == 'fulltext':
648
delta_parents = self._index.get_parents(parent)
650
if method == 'line-delta':
653
lines = self.factory.make(lines, version_id)
654
if delta or (self.factory.annotated and len(present_parents) > 0):
655
# Merge annotations from parent texts if so is needed.
656
delta_hunks = self._merge_annotations(lines, present_parents, parent_texts,
657
delta, self.factory.annotated)
793
if type(element) != str:
794
raise TypeError("key contains non-strings: %r" % (key,))
795
# Knit hunks are still last-element only
797
content = self._factory.make(lines, version_id)
798
if 'no-eol' in options:
799
# Hint to the content object that its text() call should strip the
801
content._should_strip_eol = True
802
if delta or (self._factory.annotated and len(present_parents) > 0):
803
# Merge annotations from parent texts if needed.
804
delta_hunks = self._merge_annotations(content, present_parents,
805
parent_texts, delta, self._factory.annotated,
806
left_matching_blocks)
660
809
options.append('line-delta')
661
store_lines = self.factory.lower_line_delta(delta_hunks)
810
store_lines = self._factory.lower_line_delta(delta_hunks)
811
size, bytes = self._record_to_data(key, digest,
663
814
options.append('fulltext')
664
store_lines = self.factory.lower_fulltext(lines)
666
where, size = self._data.add_record(version_id, digest, store_lines)
667
self._index.add_version(version_id, options, where, size, parents)
815
# isinstance is slower and we have no hierarchy.
816
if self._factory.__class__ == KnitPlainFactory:
817
# Use the already joined bytes saving iteration time in
819
size, bytes = self._record_to_data(key, digest,
822
# get mixed annotation + content and feed it into the
824
store_lines = self._factory.lower_fulltext(content)
825
size, bytes = self._record_to_data(key, digest,
828
access_memo = self._access.add_raw_records([(key, size)], bytes)[0]
829
self._index.add_records(
830
((key, options, access_memo, parents),),
832
return digest, text_length, content
834
def annotate(self, key):
835
"""See VersionedFiles.annotate."""
836
return self._factory.annotate(self, key)
670
838
def check(self, progress_bar=None):
671
"""See VersionedFile.check()."""
673
def _clone_text(self, new_version_id, old_version_id, parents):
674
"""See VersionedFile.clone_text()."""
675
# FIXME RBC 20060228 make fast by only inserting an index with null
677
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
679
def get_lines(self, version_id):
680
"""See VersionedFile.get_lines()."""
681
return self.get_line_list([version_id])[0]
683
def _get_record_map(self, version_ids):
684
"""Produce a dictionary of knit records.
839
"""See VersionedFiles.check()."""
840
# This doesn't actually test extraction of everything, but that will
841
# impact 'bzr check' substantially, and needs to be integrated with
842
# care. However, it does check for the obvious problem of a delta with
844
keys = self._index.keys()
845
parent_map = self.get_parent_map(keys)
847
if self._index.get_method(key) != 'fulltext':
848
compression_parent = parent_map[key][0]
849
if compression_parent not in parent_map:
850
raise errors.KnitCorrupt(self,
851
"Missing basis parent %s for %s" % (
852
compression_parent, key))
853
for fallback_vfs in self._fallback_vfs:
856
def _check_add(self, key, lines, random_id, check_content):
857
"""check that version_id and lines are safe to add."""
859
if contains_whitespace(version_id):
860
raise InvalidRevisionId(version_id, self)
861
self.check_not_reserved_id(version_id)
862
# TODO: If random_id==False and the key is already present, we should
863
# probably check that the existing content is identical to what is
864
# being inserted, and otherwise raise an exception. This would make
865
# the bundle code simpler.
867
self._check_lines_not_unicode(lines)
868
self._check_lines_are_lines(lines)
870
def _check_header(self, key, line):
871
rec = self._split_header(line)
872
self._check_header_version(rec, key[-1])
875
def _check_header_version(self, rec, version_id):
876
"""Checks the header version on original format knit records.
686
The keys are version_ids, the values are tuples of (method, content,
688
method is the way the content should be applied.
689
content is a KnitContent object.
690
digest is the SHA1 digest of this version id after all steps are done
691
next is the build-parent of the version, i.e. the leftmost ancestor.
692
If the method is fulltext, next will be None.
694
position_map = self._get_components_positions(version_ids)
695
# c = component_id, m = method, p = position, s = size, n = next
696
records = [(c, p, s) for c, (m, p, s, n) in position_map.iteritems()]
698
for component_id, content, digest in\
699
self._data.read_records_iter(records):
700
method, position, size, next = position_map[component_id]
701
record_map[component_id] = method, content, digest, next
705
def get_text(self, version_id):
706
"""See VersionedFile.get_text"""
707
return self.get_texts([version_id])[0]
709
def get_texts(self, version_ids):
710
return [''.join(l) for l in self.get_line_list(version_ids)]
712
def get_line_list(self, version_ids):
713
"""Return the texts of listed versions as a list of strings."""
714
text_map, content_map = self._get_content_maps(version_ids)
715
return [text_map[v] for v in version_ids]
717
def _get_content_maps(self, version_ids):
878
These have the last component of the key embedded in the record.
880
if rec[1] != version_id:
881
raise KnitCorrupt(self,
882
'unexpected version, wanted %r, got %r' % (version_id, rec[1]))
884
def _check_should_delta(self, parent):
885
"""Iterate back through the parent listing, looking for a fulltext.
887
This is used when we want to decide whether to add a delta or a new
888
fulltext. It searches for _max_delta_chain parents. When it finds a
889
fulltext parent, it sees if the total size of the deltas leading up to
890
it is large enough to indicate that we want a new full text anyway.
892
Return True if we should create a new delta, False if we should use a
897
for count in xrange(self._max_delta_chain):
898
# XXX: Collapse these two queries:
900
# Note that this only looks in the index of this particular
901
# KnitVersionedFiles, not in the fallbacks. This ensures that
902
# we won't store a delta spanning physical repository
904
method = self._index.get_method(parent)
905
except RevisionNotPresent:
906
# Some basis is not locally present: always delta
908
index, pos, size = self._index.get_position(parent)
909
if method == 'fulltext':
913
# We don't explicitly check for presence because this is in an
914
# inner loop, and if it's missing it'll fail anyhow.
915
# TODO: This should be asking for compression parent, not graph
917
parent = self._index.get_parent_map([parent])[parent][0]
919
# We couldn't find a fulltext, so we must create a new one
921
# Simple heuristic - if the total I/O wold be greater as a delta than
922
# the originally installed fulltext, we create a new fulltext.
923
return fulltext_size > delta_size
925
def _build_details_to_components(self, build_details):
926
"""Convert a build_details tuple to a position tuple."""
927
# record_details, access_memo, compression_parent
928
return build_details[3], build_details[0], build_details[1]
930
def _get_components_positions(self, keys, allow_missing=False):
931
"""Produce a map of position data for the components of keys.
933
This data is intended to be used for retrieving the knit records.
935
A dict of key to (record_details, index_memo, next, parents) is
937
method is the way referenced data should be applied.
938
index_memo is the handle to pass to the data access to actually get the
940
next is the build-parent of the version, or None for fulltexts.
941
parents is the version_ids of the parents of this version
943
:param allow_missing: If True do not raise an error on a missing component,
947
pending_components = keys
948
while pending_components:
949
build_details = self._index.get_build_details(pending_components)
950
current_components = set(pending_components)
951
pending_components = set()
952
for key, details in build_details.iteritems():
953
(index_memo, compression_parent, parents,
954
record_details) = details
955
method = record_details[0]
956
if compression_parent is not None:
957
pending_components.add(compression_parent)
958
component_data[key] = self._build_details_to_components(details)
959
missing = current_components.difference(build_details)
960
if missing and not allow_missing:
961
raise errors.RevisionNotPresent(missing.pop(), self)
962
return component_data
964
def _get_content(self, key, parent_texts={}):
965
"""Returns a content object that makes up the specified
967
cached_version = parent_texts.get(key, None)
968
if cached_version is not None:
969
# Ensure the cache dict is valid.
970
if not self.get_parent_map([key]):
971
raise RevisionNotPresent(key, self)
972
return cached_version
973
text_map, contents_map = self._get_content_maps([key])
974
return contents_map[key]
976
def _get_content_maps(self, keys, nonlocal_keys=None):
718
977
"""Produce maps of text and KnitContents
979
:param keys: The keys to produce content maps for.
980
:param nonlocal_keys: An iterable of keys(possibly intersecting keys)
981
which are known to not be in this knit, but rather in one of the
720
983
:return: (text_map, content_map) where text_map contains the texts for
721
the requested versions and content_map contains the KnitContents.
722
Both dicts take version_ids as their keys.
984
the requested versions and content_map contains the KnitContents.
724
for version_id in version_ids:
725
if not self.has_version(version_id):
726
raise RevisionNotPresent(version_id, self.filename)
727
record_map = self._get_record_map(version_ids)
986
# FUTURE: This function could be improved for the 'extract many' case
987
# by tracking each component and only doing the copy when the number of
988
# children than need to apply delta's to it is > 1 or it is part of the
991
multiple_versions = len(keys) != 1
992
record_map = self._get_record_map(keys, allow_missing=True)
731
996
final_content = {}
732
for version_id in version_ids:
997
if nonlocal_keys is None:
998
nonlocal_keys = set()
1000
nonlocal_keys = frozenset(nonlocal_keys)
1001
missing_keys = set(nonlocal_keys)
1002
for source in self._fallback_vfs:
1003
if not missing_keys:
1005
for record in source.get_record_stream(missing_keys,
1007
if record.storage_kind == 'absent':
1009
missing_keys.remove(record.key)
1010
lines = split_lines(record.get_bytes_as('fulltext'))
1011
text_map[record.key] = lines
1012
content_map[record.key] = PlainKnitContent(lines, record.key)
1013
if record.key in keys:
1014
final_content[record.key] = content_map[record.key]
1016
if key in nonlocal_keys:
735
1021
while cursor is not None:
736
method, data, digest, next = record_map[cursor]
737
components.append((cursor, method, data, digest))
1023
record, record_details, digest, next = record_map[cursor]
1025
raise RevisionNotPresent(cursor, self)
1026
components.append((cursor, record, record_details, digest))
738
1028
if cursor in content_map:
1029
# no need to plan further back
1030
components.append((cursor, None, None, None))
743
for component_id, method, data, digest in reversed(components):
1034
for (component_id, record, record_details,
1035
digest) in reversed(components):
744
1036
if component_id in content_map:
745
1037
content = content_map[component_id]
747
version_idx = self._index.lookup(component_id)
748
if method == 'fulltext':
749
assert content is None
750
content = self.factory.parse_fulltext(data, version_idx)
751
elif method == 'line-delta':
752
delta = self.factory.parse_line_delta(data[:],
754
content = content.copy()
755
content._lines = self._apply_delta(content._lines,
757
content_map[component_id] = content
1039
content, delta = self._factory.parse_record(key[-1],
1040
record, record_details, content,
1041
copy_base_content=multiple_versions)
1042
if multiple_versions:
1043
content_map[component_id] = content
759
if 'no-eol' in self._index.get_options(version_id):
760
content = content.copy()
761
line = content._lines[-1][1].rstrip('\n')
762
content._lines[-1] = (content._lines[-1][0], line)
763
final_content[version_id] = content
1045
final_content[key] = content
765
1047
# digest here is the digest from the last applied component.
766
1048
text = content.text()
767
if sha_strings(text) != digest:
768
raise KnitCorrupt(self.filename,
769
'sha-1 does not match %s' % version_id)
771
text_map[version_id] = text
772
return text_map, final_content
774
def iter_lines_added_or_present_in_versions(self, version_ids=None):
775
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
776
if version_ids is None:
777
version_ids = self.versions()
1049
actual_sha = sha_strings(text)
1050
if actual_sha != digest:
1051
raise KnitCorrupt(self,
1053
'\n of reconstructed text does not match'
1055
'\n for version %s' %
1056
(actual_sha, digest, key))
1057
text_map[key] = text
1058
return text_map, final_content
1060
def get_parent_map(self, keys):
1061
"""Get a map of the graph parents of keys.
1063
:param keys: The keys to look up parents for.
1064
:return: A mapping from keys to parents. Absent keys are absent from
1067
return self._get_parent_map_with_sources(keys)[0]
1069
def _get_parent_map_with_sources(self, keys):
1070
"""Get a map of the parents of keys.
1072
:param keys: The keys to look up parents for.
1073
:return: A tuple. The first element is a mapping from keys to parents.
1074
Absent keys are absent from the mapping. The second element is a
1075
list with the locations each key was found in. The first element
1076
is the in-this-knit parents, the second the first fallback source,
1080
sources = [self._index] + self._fallback_vfs
1083
for source in sources:
1086
new_result = source.get_parent_map(missing)
1087
source_results.append(new_result)
1088
result.update(new_result)
1089
missing.difference_update(set(new_result))
1090
return result, source_results
1092
def _get_record_map(self, keys, allow_missing=False):
1093
"""Produce a dictionary of knit records.
1095
:return: {key:(record, record_details, digest, next)}
1097
data returned from read_records
1099
opaque information to pass to parse_record
1101
SHA1 digest of the full text after all steps are done
1103
build-parent of the version, i.e. the leftmost ancestor.
1104
Will be None if the record is not a delta.
1105
:param keys: The keys to build a map for
1106
:param allow_missing: If some records are missing, rather than
1107
error, just return the data that could be generated.
1109
position_map = self._get_components_positions(keys,
1110
allow_missing=allow_missing)
1111
# key = component_id, r = record_details, i_m = index_memo, n = next
1112
records = [(key, i_m) for key, (r, i_m, n)
1113
in position_map.iteritems()]
1115
for key, record, digest in \
1116
self._read_records_iter(records):
1117
(record_details, index_memo, next) = position_map[key]
1118
record_map[key] = record, record_details, digest, next
1121
def get_record_stream(self, keys, ordering, include_delta_closure):
1122
"""Get a stream of records for keys.
1124
:param keys: The keys to include.
1125
:param ordering: Either 'unordered' or 'topological'. A topologically
1126
sorted stream has compression parents strictly before their
1128
:param include_delta_closure: If True then the closure across any
1129
compression parents will be included (in the opaque data).
1130
:return: An iterator of ContentFactory objects, each of which is only
1131
valid until the iterator is advanced.
1133
# keys might be a generator
1137
if not self._index.has_graph:
1138
# Cannot topological order when no graph has been stored.
1139
ordering = 'unordered'
1140
if include_delta_closure:
1141
positions = self._get_components_positions(keys, allow_missing=True)
1143
build_details = self._index.get_build_details(keys)
1145
# (record_details, access_memo, compression_parent_key)
1146
positions = dict((key, self._build_details_to_components(details))
1147
for key, details in build_details.iteritems())
1148
absent_keys = keys.difference(set(positions))
1149
# There may be more absent keys : if we're missing the basis component
1150
# and are trying to include the delta closure.
1151
if include_delta_closure:
1152
needed_from_fallback = set()
1153
# Build up reconstructable_keys dict. key:True in this dict means
1154
# the key can be reconstructed.
1155
reconstructable_keys = {}
1159
chain = [key, positions[key][2]]
1161
needed_from_fallback.add(key)
1164
while chain[-1] is not None:
1165
if chain[-1] in reconstructable_keys:
1166
result = reconstructable_keys[chain[-1]]
1170
chain.append(positions[chain[-1]][2])
1172
# missing basis component
1173
needed_from_fallback.add(chain[-1])
1176
for chain_key in chain[:-1]:
1177
reconstructable_keys[chain_key] = result
1179
needed_from_fallback.add(key)
1180
# Double index lookups here : need a unified api ?
1181
global_map, parent_maps = self._get_parent_map_with_sources(keys)
1182
if ordering == 'topological':
1183
# Global topological sort
1184
present_keys = tsort.topo_sort(global_map)
1185
# Now group by source:
1187
current_source = None
1188
for key in present_keys:
1189
for parent_map in parent_maps:
1190
if key in parent_map:
1191
key_source = parent_map
1193
if current_source is not key_source:
1194
source_keys.append((key_source, []))
1195
current_source = key_source
1196
source_keys[-1][1].append(key)
1198
# Just group by source; remote sources first.
1201
for parent_map in reversed(parent_maps):
1202
source_keys.append((parent_map, []))
1203
for key in parent_map:
1204
present_keys.append(key)
1205
source_keys[-1][1].append(key)
1206
absent_keys = keys - set(global_map)
1207
for key in absent_keys:
1208
yield AbsentContentFactory(key)
1209
# restrict our view to the keys we can answer.
1210
# XXX: Memory: TODO: batch data here to cap buffered data at (say) 1MB.
1211
# XXX: At that point we need to consider the impact of double reads by
1212
# utilising components multiple times.
1213
if include_delta_closure:
1214
# XXX: get_content_maps performs its own index queries; allow state
1216
text_map, _ = self._get_content_maps(present_keys,
1217
needed_from_fallback - absent_keys)
1218
for key in present_keys:
1219
yield FulltextContentFactory(key, global_map[key], None,
1220
''.join(text_map[key]))
1222
for source, keys in source_keys:
1223
if source is parent_maps[0]:
1224
# this KnitVersionedFiles
1225
records = [(key, positions[key][1]) for key in keys]
1226
for key, raw_data, sha1 in self._read_records_iter_raw(records):
1227
(record_details, index_memo, _) = positions[key]
1228
yield KnitContentFactory(key, global_map[key],
1229
record_details, sha1, raw_data, self._factory.annotated, None)
1231
vf = self._fallback_vfs[parent_maps.index(source) - 1]
1232
for record in vf.get_record_stream(keys, ordering,
1233
include_delta_closure):
1236
def get_sha1s(self, keys):
1237
"""See VersionedFiles.get_sha1s()."""
1239
record_map = self._get_record_map(missing, allow_missing=True)
1241
for key, details in record_map.iteritems():
1242
if key not in missing:
1244
# record entry 2 is the 'digest'.
1245
result[key] = details[2]
1246
missing.difference_update(set(result))
1247
for source in self._fallback_vfs:
1250
new_result = source.get_sha1s(missing)
1251
result.update(new_result)
1252
missing.difference_update(set(new_result))
1255
def insert_record_stream(self, stream):
1256
"""Insert a record stream into this container.
1258
:param stream: A stream of records to insert.
1260
:seealso VersionedFiles.get_record_stream:
1262
def get_adapter(adapter_key):
1264
return adapters[adapter_key]
1266
adapter_factory = adapter_registry.get(adapter_key)
1267
adapter = adapter_factory(self)
1268
adapters[adapter_key] = adapter
1270
if self._factory.annotated:
1271
# self is annotated, we need annotated knits to use directly.
1272
annotated = "annotated-"
1275
# self is not annotated, but we can strip annotations cheaply.
1277
convertibles = set(["knit-annotated-ft-gz"])
1278
if self._max_delta_chain:
1279
convertibles.add("knit-annotated-delta-gz")
1280
# The set of types we can cheaply adapt without needing basis texts.
1281
native_types = set()
1282
if self._max_delta_chain:
1283
native_types.add("knit-%sdelta-gz" % annotated)
1284
native_types.add("knit-%sft-gz" % annotated)
1285
knit_types = native_types.union(convertibles)
1287
# Buffer all index entries that we can't add immediately because their
1288
# basis parent is missing. We don't buffer all because generating
1289
# annotations may require access to some of the new records. However we
1290
# can't generate annotations from new deltas until their basis parent
1291
# is present anyway, so we get away with not needing an index that
1292
# includes the new keys.
1293
# key = basis_parent, value = index entry to add
1294
buffered_index_entries = {}
1295
for record in stream:
1296
parents = record.parents
1297
# Raise an error when a record is missing.
1298
if record.storage_kind == 'absent':
1299
raise RevisionNotPresent([record.key], self)
1300
if record.storage_kind in knit_types:
1301
if record.storage_kind not in native_types:
1303
adapter_key = (record.storage_kind, "knit-delta-gz")
1304
adapter = get_adapter(adapter_key)
1306
adapter_key = (record.storage_kind, "knit-ft-gz")
1307
adapter = get_adapter(adapter_key)
1308
bytes = adapter.get_bytes(
1309
record, record.get_bytes_as(record.storage_kind))
1311
bytes = record.get_bytes_as(record.storage_kind)
1312
options = [record._build_details[0]]
1313
if record._build_details[1]:
1314
options.append('no-eol')
1315
# Just blat it across.
1316
# Note: This does end up adding data on duplicate keys. As
1317
# modern repositories use atomic insertions this should not
1318
# lead to excessive growth in the event of interrupted fetches.
1319
# 'knit' repositories may suffer excessive growth, but as a
1320
# deprecated format this is tolerable. It can be fixed if
1321
# needed by in the kndx index support raising on a duplicate
1322
# add with identical parents and options.
1323
access_memo = self._access.add_raw_records(
1324
[(record.key, len(bytes))], bytes)[0]
1325
index_entry = (record.key, options, access_memo, parents)
1327
if 'fulltext' not in options:
1328
basis_parent = parents[0]
1329
# Note that pack backed knits don't need to buffer here
1330
# because they buffer all writes to the transaction level,
1331
# but we don't expose that difference at the index level. If
1332
# the query here has sufficient cost to show up in
1333
# profiling we should do that.
1334
if basis_parent not in self.get_parent_map([basis_parent]):
1335
pending = buffered_index_entries.setdefault(
1337
pending.append(index_entry)
1340
self._index.add_records([index_entry])
1341
elif record.storage_kind == 'fulltext':
1342
self.add_lines(record.key, parents,
1343
split_lines(record.get_bytes_as('fulltext')))
1345
adapter_key = record.storage_kind, 'fulltext'
1346
adapter = get_adapter(adapter_key)
1347
lines = split_lines(adapter.get_bytes(
1348
record, record.get_bytes_as(record.storage_kind)))
1350
self.add_lines(record.key, parents, lines)
1351
except errors.RevisionAlreadyPresent:
1353
# Add any records whose basis parent is now available.
1354
added_keys = [record.key]
1356
key = added_keys.pop(0)
1357
if key in buffered_index_entries:
1358
index_entries = buffered_index_entries[key]
1359
self._index.add_records(index_entries)
1361
[index_entry[0] for index_entry in index_entries])
1362
del buffered_index_entries[key]
1363
# If there were any deltas which had a missing basis parent, error.
1364
if buffered_index_entries:
1365
raise errors.RevisionNotPresent(buffered_index_entries.keys()[0],
1368
def iter_lines_added_or_present_in_keys(self, keys, pb=None):
1369
"""Iterate over the lines in the versioned files from keys.
1371
This may return lines from other keys. Each item the returned
1372
iterator yields is a tuple of a line and a text version that that line
1373
is present in (not introduced in).
1375
Ordering of results is in whatever order is most suitable for the
1376
underlying storage format.
1378
If a progress bar is supplied, it may be used to indicate progress.
1379
The caller is responsible for cleaning up progress bars (because this
1383
* Lines are normalised by the underlying store: they will all have \n
1385
* Lines are returned in arbitrary order.
1387
:return: An iterator over (line, key).
1390
pb = progress.DummyProgress()
778
1393
# we don't care about inclusions, the caller cares.
779
1394
# but we need to setup a list of records to visit.
780
# we need version_id, position, length
781
version_id_records = []
782
requested_versions = list(version_ids)
783
# filter for available versions
784
for version_id in requested_versions:
785
if not self.has_version(version_id):
786
raise RevisionNotPresent(version_id, self.filename)
787
# get a in-component-order queue:
789
for version_id in self.versions():
790
if version_id in requested_versions:
791
version_ids.append(version_id)
792
data_pos, length = self._index.get_position(version_id)
793
version_id_records.append((version_id, data_pos, length))
795
pb = bzrlib.ui.ui_factory.nested_progress_bar()
797
total = len(version_id_records)
799
pb.update('Walking content.', count, total)
800
for version_id, data, sha_value in \
801
self._data.read_records_iter(version_id_records):
802
pb.update('Walking content.', count, total)
803
method = self._index.get_method(version_id)
804
version_idx = self._index.lookup(version_id)
805
assert method in ('fulltext', 'line-delta')
806
if method == 'fulltext':
807
content = self.factory.parse_fulltext(data, version_idx)
808
for line in content.text():
1395
# we need key, position, length
1397
build_details = self._index.get_build_details(keys)
1398
for key, details in build_details.iteritems():
1400
key_records.append((key, details[0]))
1402
records_iter = enumerate(self._read_records_iter(key_records))
1403
for (key_idx, (key, data, sha_value)) in records_iter:
1404
pb.update('Walking content.', key_idx, total)
1405
compression_parent = build_details[key][1]
1406
if compression_parent is None:
1408
line_iterator = self._factory.get_fulltext_content(data)
1411
line_iterator = self._factory.get_linedelta_content(data)
1412
# XXX: It might be more efficient to yield (key,
1413
# line_iterator) in the future. However for now, this is a simpler
1414
# change to integrate into the rest of the codebase. RBC 20071110
1415
for line in line_iterator:
1417
for source in self._fallback_vfs:
1421
for line, key in source.iter_lines_added_or_present_in_keys(keys):
1422
source_keys.add(key)
1424
keys.difference_update(source_keys)
1426
raise RevisionNotPresent(keys, self.filename)
1427
pb.update('Walking content.', total, total)
1429
def _make_line_delta(self, delta_seq, new_content):
1430
"""Generate a line delta from delta_seq and new_content."""
1432
for op in delta_seq.get_opcodes():
1433
if op[0] == 'equal':
1435
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
1438
def _merge_annotations(self, content, parents, parent_texts={},
1439
delta=None, annotated=None,
1440
left_matching_blocks=None):
1441
"""Merge annotations for content and generate deltas.
1443
This is done by comparing the annotations based on changes to the text
1444
and generating a delta on the resulting full texts. If annotations are
1445
not being created then a simple delta is created.
1447
if left_matching_blocks is not None:
1448
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
1452
for parent_key in parents:
1453
merge_content = self._get_content(parent_key, parent_texts)
1454
if (parent_key == parents[0] and delta_seq is not None):
811
delta = self.factory.parse_line_delta(data, version_idx)
812
for start, end, count, lines in delta:
813
for origin, line in lines:
816
pb.update('Walking content.', total, total)
819
pb.update('Walking content.', total, total)
823
def num_versions(self):
824
"""See VersionedFile.num_versions()."""
825
return self._index.num_versions()
827
__len__ = num_versions
829
def annotate_iter(self, version_id):
830
"""See VersionedFile.annotate_iter."""
831
content = self._get_content(version_id)
832
for origin, text in content.annotate_iter():
835
def get_parents(self, version_id):
836
"""See VersionedFile.get_parents."""
839
# 52554 calls in 1264 872 internal down from 3674
841
return self._index.get_parents(version_id)
843
raise RevisionNotPresent(version_id, self.filename)
845
def get_parents_with_ghosts(self, version_id):
846
"""See VersionedFile.get_parents."""
848
return self._index.get_parents_with_ghosts(version_id)
850
raise RevisionNotPresent(version_id, self.filename)
852
def get_ancestry(self, versions):
853
"""See VersionedFile.get_ancestry."""
854
if isinstance(versions, basestring):
855
versions = [versions]
858
self._check_versions_present(versions)
859
return self._index.get_ancestry(versions)
861
def get_ancestry_with_ghosts(self, versions):
862
"""See VersionedFile.get_ancestry_with_ghosts."""
863
if isinstance(versions, basestring):
864
versions = [versions]
867
self._check_versions_present(versions)
868
return self._index.get_ancestry_with_ghosts(versions)
870
#@deprecated_method(zero_eight)
871
def walk(self, version_ids):
872
"""See VersionedFile.walk."""
873
# We take the short path here, and extract all relevant texts
874
# and put them in a weave and let that do all the work. Far
875
# from optimal, but is much simpler.
876
# FIXME RB 20060228 this really is inefficient!
877
from bzrlib.weave import Weave
879
w = Weave(self.filename)
880
ancestry = self.get_ancestry(version_ids)
881
sorted_graph = topo_sort(self._index.get_graph())
882
version_list = [vid for vid in sorted_graph if vid in ancestry]
884
for version_id in version_list:
885
lines = self.get_lines(version_id)
886
w.add_lines(version_id, self.get_parents(version_id), lines)
888
for lineno, insert_id, dset, line in w.walk(version_ids):
889
yield lineno, insert_id, dset, line
891
def plan_merge(self, ver_a, ver_b):
892
"""See VersionedFile.plan_merge."""
893
ancestors_b = set(self.get_ancestry(ver_b))
894
def status_a(revision, text):
895
if revision in ancestors_b:
896
return 'killed-b', text
900
ancestors_a = set(self.get_ancestry(ver_a))
901
def status_b(revision, text):
902
if revision in ancestors_a:
903
return 'killed-a', text
907
annotated_a = self.annotate(ver_a)
908
annotated_b = self.annotate(ver_b)
909
plain_a = [t for (a, t) in annotated_a]
910
plain_b = [t for (a, t) in annotated_b]
911
blocks = KnitSequenceMatcher(None, plain_a, plain_b).get_matching_blocks()
914
for ai, bi, l in blocks:
915
# process all mismatched sections
916
# (last mismatched section is handled because blocks always
917
# includes a 0-length last block)
918
for revision, text in annotated_a[a_cur:ai]:
919
yield status_a(revision, text)
920
for revision, text in annotated_b[b_cur:bi]:
921
yield status_b(revision, text)
923
# and now the matched section
926
for text_a, text_b in zip(plain_a[ai:a_cur], plain_b[bi:b_cur]):
927
assert text_a == text_b
928
yield "unchanged", text_a
931
class _KnitComponentFile(object):
932
"""One of the files used to implement a knit database"""
934
def __init__(self, transport, filename, mode, file_mode=None):
935
self._transport = transport
936
self._filename = filename
938
self._file_mode=file_mode
940
def write_header(self):
941
if self._transport.append(self._filename, StringIO(self.HEADER),
942
mode=self._file_mode):
943
raise KnitCorrupt(self._filename, 'misaligned after writing header')
945
def check_header(self, fp):
947
if line != self.HEADER:
948
raise KnitHeaderError(badline=line)
951
"""Commit is a nop."""
954
return '%s(%s)' % (self.__class__.__name__, self._filename)
957
class _KnitIndex(_KnitComponentFile):
958
"""Manages knit index file.
960
The index is already kept in memory and read on startup, to enable
1457
seq = patiencediff.PatienceSequenceMatcher(
1458
None, merge_content.text(), content.text())
1459
for i, j, n in seq.get_matching_blocks():
1462
# this copies (origin, text) pairs across to the new
1463
# content for any line that matches the last-checked
1465
content._lines[j:j+n] = merge_content._lines[i:i+n]
1466
# XXX: Robert says the following block is a workaround for a
1467
# now-fixed bug and it can probably be deleted. -- mbp 20080618
1468
if content._lines and content._lines[-1][1][-1] != '\n':
1469
# The copied annotation was from a line without a trailing EOL,
1470
# reinstate one for the content object, to ensure correct
1472
line = content._lines[-1][1] + '\n'
1473
content._lines[-1] = (content._lines[-1][0], line)
1475
if delta_seq is None:
1476
reference_content = self._get_content(parents[0], parent_texts)
1477
new_texts = content.text()
1478
old_texts = reference_content.text()
1479
delta_seq = patiencediff.PatienceSequenceMatcher(
1480
None, old_texts, new_texts)
1481
return self._make_line_delta(delta_seq, content)
1483
def _parse_record(self, version_id, data):
1484
"""Parse an original format knit record.
1486
These have the last element of the key only present in the stored data.
1488
rec, record_contents = self._parse_record_unchecked(data)
1489
self._check_header_version(rec, version_id)
1490
return record_contents, rec[3]
1492
def _parse_record_header(self, key, raw_data):
1493
"""Parse a record header for consistency.
1495
:return: the header and the decompressor stream.
1496
as (stream, header_record)
1498
df = tuned_gzip.GzipFile(mode='rb', fileobj=StringIO(raw_data))
1501
rec = self._check_header(key, df.readline())
1502
except Exception, e:
1503
raise KnitCorrupt(self,
1504
"While reading {%s} got %s(%s)"
1505
% (key, e.__class__.__name__, str(e)))
1508
def _parse_record_unchecked(self, data):
1510
# 4168 calls in 2880 217 internal
1511
# 4168 calls to _parse_record_header in 2121
1512
# 4168 calls to readlines in 330
1513
df = tuned_gzip.GzipFile(mode='rb', fileobj=StringIO(data))
1515
record_contents = df.readlines()
1516
except Exception, e:
1517
raise KnitCorrupt(self, "Corrupt compressed record %r, got %s(%s)" %
1518
(data, e.__class__.__name__, str(e)))
1519
header = record_contents.pop(0)
1520
rec = self._split_header(header)
1521
last_line = record_contents.pop()
1522
if len(record_contents) != int(rec[2]):
1523
raise KnitCorrupt(self,
1524
'incorrect number of lines %s != %s'
1525
' for version {%s} %s'
1526
% (len(record_contents), int(rec[2]),
1527
rec[1], record_contents))
1528
if last_line != 'end %s\n' % rec[1]:
1529
raise KnitCorrupt(self,
1530
'unexpected version end line %r, wanted %r'
1531
% (last_line, rec[1]))
1533
return rec, record_contents
1535
def _read_records_iter(self, records):
1536
"""Read text records from data file and yield result.
1538
The result will be returned in whatever is the fastest to read.
1539
Not by the order requested. Also, multiple requests for the same
1540
record will only yield 1 response.
1541
:param records: A list of (key, access_memo) entries
1542
:return: Yields (key, contents, digest) in the order
1543
read, not the order requested
1548
# XXX: This smells wrong, IO may not be getting ordered right.
1549
needed_records = sorted(set(records), key=operator.itemgetter(1))
1550
if not needed_records:
1553
# The transport optimizes the fetching as well
1554
# (ie, reads continuous ranges.)
1555
raw_data = self._access.get_raw_records(
1556
[index_memo for key, index_memo in needed_records])
1558
for (key, index_memo), data in \
1559
izip(iter(needed_records), raw_data):
1560
content, digest = self._parse_record(key[-1], data)
1561
yield key, content, digest
1563
def _read_records_iter_raw(self, records):
1564
"""Read text records from data file and yield raw data.
1566
This unpacks enough of the text record to validate the id is
1567
as expected but thats all.
1569
Each item the iterator yields is (key, bytes, sha1_of_full_text).
1571
# setup an iterator of the external records:
1572
# uses readv so nice and fast we hope.
1574
# grab the disk data needed.
1575
needed_offsets = [index_memo for key, index_memo
1577
raw_records = self._access.get_raw_records(needed_offsets)
1579
for key, index_memo in records:
1580
data = raw_records.next()
1581
# validate the header (note that we can only use the suffix in
1582
# current knit records).
1583
df, rec = self._parse_record_header(key, data)
1585
yield key, data, rec[3]
1587
def _record_to_data(self, key, digest, lines, dense_lines=None):
1588
"""Convert key, digest, lines into a raw data block.
1590
:param key: The key of the record. Currently keys are always serialised
1591
using just the trailing component.
1592
:param dense_lines: The bytes of lines but in a denser form. For
1593
instance, if lines is a list of 1000 bytestrings each ending in \n,
1594
dense_lines may be a list with one line in it, containing all the
1595
1000's lines and their \n's. Using dense_lines if it is already
1596
known is a win because the string join to create bytes in this
1597
function spends less time resizing the final string.
1598
:return: (len, a StringIO instance with the raw data ready to read.)
1600
# Note: using a string copy here increases memory pressure with e.g.
1601
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
1602
# when doing the initial commit of a mozilla tree. RBC 20070921
1603
bytes = ''.join(chain(
1604
["version %s %d %s\n" % (key[-1],
1607
dense_lines or lines,
1608
["end %s\n" % key[-1]]))
1609
if type(bytes) != str:
1610
raise AssertionError(
1611
'data must be plain bytes was %s' % type(bytes))
1612
if lines and lines[-1][-1] != '\n':
1613
raise ValueError('corrupt lines value %r' % lines)
1614
compressed_bytes = tuned_gzip.bytes_to_gzip(bytes)
1615
return len(compressed_bytes), compressed_bytes
1617
def _split_header(self, line):
1620
raise KnitCorrupt(self,
1621
'unexpected number of elements in record header')
1625
"""See VersionedFiles.keys."""
1626
if 'evil' in debug.debug_flags:
1627
trace.mutter_callsite(2, "keys scales with size of history")
1628
sources = [self._index] + self._fallback_vfs
1630
for source in sources:
1631
result.update(source.keys())
1636
class _KndxIndex(object):
1637
"""Manages knit index files
1639
The index is kept in memory and read on startup, to enable
961
1640
fast lookups of revision information. The cursor of the index
962
1641
file is always pointing to the end, making it easy to append
996
1675
The ' :' marker is the end of record marker.
999
when a write is interrupted to the index file, it will result in a line that
1000
does not end in ' :'. If the ' :' is not present at the end of a line, or at
1001
the end of the file, then the record that is missing it will be ignored by
1678
when a write is interrupted to the index file, it will result in a line
1679
that does not end in ' :'. If the ' :' is not present at the end of a line,
1680
or at the end of the file, then the record that is missing it will be
1681
ignored by the parser.
1004
1683
When writing new records to the index file, the data is preceded by '\n'
1005
1684
to ensure that records always start on new lines even if the last write was
1006
1685
interrupted. As a result its normal for the last line in the index to be
1007
1686
missing a trailing newline. One can be added with no harmful effects.
1688
:ivar _kndx_cache: dict from prefix to the old state of KnitIndex objects,
1689
where prefix is e.g. the (fileid,) for .texts instances or () for
1690
constant-mapped things like .revisions, and the old state is
1691
tuple(history_vector, cache_dict). This is used to prevent having an
1692
ABI change with the C extension that reads .kndx files.
1010
1695
HEADER = "# bzr knit index 8\n"
1012
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1013
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1015
def _cache_version(self, version_id, options, pos, size, parents):
1697
def __init__(self, transport, mapper, get_scope, allow_writes, is_locked):
1698
"""Create a _KndxIndex on transport using mapper."""
1699
self._transport = transport
1700
self._mapper = mapper
1701
self._get_scope = get_scope
1702
self._allow_writes = allow_writes
1703
self._is_locked = is_locked
1705
self.has_graph = True
1707
def add_records(self, records, random_id=False):
1708
"""Add multiple records to the index.
1710
:param records: a list of tuples:
1711
(key, options, access_memo, parents).
1712
:param random_id: If True the ids being added were randomly generated
1713
and no check for existence will be performed.
1716
for record in records:
1719
path = self._mapper.map(key) + '.kndx'
1720
path_keys = paths.setdefault(path, (prefix, []))
1721
path_keys[1].append(record)
1722
for path in sorted(paths):
1723
prefix, path_keys = paths[path]
1724
self._load_prefixes([prefix])
1726
orig_history = self._kndx_cache[prefix][1][:]
1727
orig_cache = self._kndx_cache[prefix][0].copy()
1730
for key, options, (_, pos, size), parents in path_keys:
1732
# kndx indices cannot be parentless.
1734
line = "\n%s %s %s %s %s :" % (
1735
key[-1], ','.join(options), pos, size,
1736
self._dictionary_compress(parents))
1737
if type(line) != str:
1738
raise AssertionError(
1739
'data must be utf8 was %s' % type(line))
1741
self._cache_key(key, options, pos, size, parents)
1742
if len(orig_history):
1743
self._transport.append_bytes(path, ''.join(lines))
1745
self._init_index(path, lines)
1747
# If any problems happen, restore the original values and re-raise
1748
self._kndx_cache[prefix] = (orig_cache, orig_history)
1751
def _cache_key(self, key, options, pos, size, parent_keys):
1016
1752
"""Cache a version record in the history array and index cache.
1018
This is inlined into __init__ for performance. KEEP IN SYNC.
1754
This is inlined into _load_data for performance. KEEP IN SYNC.
1019
1755
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1759
version_id = key[-1]
1760
# last-element only for compatibilty with the C load_data.
1761
parents = tuple(parent[-1] for parent in parent_keys)
1762
for parent in parent_keys:
1763
if parent[:-1] != prefix:
1764
raise ValueError("mismatched prefixes for %r, %r" % (
1766
cache, history = self._kndx_cache[prefix]
1022
1767
# only want the _history index to reference the 1st index entry
1023
1768
# for version_id
1024
if version_id not in self._cache:
1025
index = len(self._history)
1026
self._history.append(version_id)
1769
if version_id not in cache:
1770
index = len(history)
1771
history.append(version_id)
1028
index = self._cache[version_id][5]
1029
self._cache[version_id] = (version_id,
1773
index = cache[version_id][5]
1774
cache[version_id] = (version_id,
1036
def __init__(self, transport, filename, mode, create=False, file_mode=None):
1037
_KnitComponentFile.__init__(self, transport, filename, mode, file_mode)
1039
# position in _history is the 'official' index for a revision
1040
# but the values may have come from a newer entry.
1041
# so - wc -l of a knit index is != the number of unique names
1044
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1781
def check_header(self, fp):
1782
line = fp.readline()
1784
# An empty file can actually be treated as though the file doesn't
1786
raise errors.NoSuchFile(self)
1787
if line != self.HEADER:
1788
raise KnitHeaderError(badline=line, filename=self)
1790
def _check_read(self):
1791
if not self._is_locked():
1792
raise errors.ObjectNotLocked(self)
1793
if self._get_scope() != self._scope:
1796
def _check_write_ok(self):
1797
"""Assert if not writes are permitted."""
1798
if not self._is_locked():
1799
raise errors.ObjectNotLocked(self)
1800
if self._get_scope() != self._scope:
1802
if self._mode != 'w':
1803
raise errors.ReadOnlyObjectDirtiedError(self)
1805
def get_build_details(self, keys):
1806
"""Get the method, index_memo and compression parent for keys.
1808
Ghosts are omitted from the result.
1810
:param keys: An iterable of keys.
1811
:return: A dict of key:(index_memo, compression_parent, parents,
1814
opaque structure to pass to read_records to extract the raw
1817
Content that this record is built upon, may be None
1819
Logical parents of this node
1821
extra information about the content which needs to be passed to
1822
Factory.parse_record
1824
prefixes = self._partition_keys(keys)
1825
parent_map = self.get_parent_map(keys)
1828
if key not in parent_map:
1830
method = self.get_method(key)
1831
parents = parent_map[key]
1832
if method == 'fulltext':
1833
compression_parent = None
1835
compression_parent = parents[0]
1836
noeol = 'no-eol' in self.get_options(key)
1837
index_memo = self.get_position(key)
1838
result[key] = (index_memo, compression_parent,
1839
parents, (method, noeol))
1842
def get_method(self, key):
1843
"""Return compression method of specified key."""
1844
options = self.get_options(key)
1845
if 'fulltext' in options:
1847
elif 'line-delta' in options:
1850
raise errors.KnitIndexUnknownMethod(self, options)
1852
def get_options(self, key):
1853
"""Return a list representing options.
1857
prefix, suffix = self._split_key(key)
1858
self._load_prefixes([prefix])
1049
pb.update('read knit index', count, total)
1050
fp = self._transport.get(self._filename)
1052
self.check_header(fp)
1053
# readlines reads the whole file at once:
1054
# bad for transports like http, good for local disk
1055
# we save 60 ms doing this one change (
1056
# from calling readline each time to calling
1058
# probably what we want for nice behaviour on
1059
# http is a incremental readlines that yields, or
1060
# a check for local vs non local indexes,
1061
for l in fp.readlines():
1063
if len(rec) < 5 or rec[-1] != ':':
1065
# FIXME: in the future we should determine if its a
1066
# short write - and ignore it
1067
# or a different failure, and raise. RBC 20060407
1071
#pb.update('read knit index', count, total)
1072
# See self._parse_parents
1074
for value in rec[4:-1]:
1076
# uncompressed reference
1077
parents.append(value[1:])
1079
# this is 15/4000ms faster than isinstance,
1081
# this function is called thousands of times a
1082
# second so small variations add up.
1083
assert value.__class__ is str
1084
parents.append(self._history[int(value)])
1085
# end self._parse_parents
1086
# self._cache_version(rec[0],
1087
# rec[1].split(','),
1091
# --- self._cache_version
1092
# only want the _history index to reference the 1st
1093
# index entry for version_id
1095
if version_id not in self._cache:
1096
index = len(self._history)
1097
self._history.append(version_id)
1099
index = self._cache[version_id][5]
1100
self._cache[version_id] = (version_id,
1106
# --- self._cache_version
1109
except NoSuchFile, e:
1110
if mode != 'w' or not create:
1114
pb.update('read knit index', total, total)
1117
def _parse_parents(self, compressed_parents):
1118
"""convert a list of string parent values into version ids.
1120
ints are looked up in the index.
1121
.FOO values are ghosts and converted in to FOO.
1123
NOTE: the function is retained here for clarity, and for possible
1124
use in partial index reads. However bulk processing now has
1125
it inlined in __init__ for inner-loop optimisation.
1860
return self._kndx_cache[prefix][0][suffix][1]
1862
raise RevisionNotPresent(key, self)
1864
def get_parent_map(self, keys):
1865
"""Get a map of the parents of keys.
1867
:param keys: The keys to look up parents for.
1868
:return: A mapping from keys to parents. Absent keys are absent from
1128
for value in compressed_parents:
1129
if value[-1] == '.':
1130
# uncompressed reference
1131
result.append(value[1:])
1133
# this is 15/4000ms faster than isinstance,
1134
# this function is called thousands of times a
1135
# second so small variations add up.
1136
assert value.__class__ is str
1137
result.append(self._history[int(value)])
1140
def get_graph(self):
1142
for version_id, index in self._cache.iteritems():
1143
graph.append((version_id, index[4]))
1146
def get_ancestry(self, versions):
1147
"""See VersionedFile.get_ancestry."""
1148
# get a graph of all the mentioned versions:
1150
pending = set(versions)
1152
version = pending.pop()
1153
parents = self._cache[version][4]
1154
# got the parents ok
1156
parents = [parent for parent in parents if parent in self._cache]
1157
for parent in parents:
1158
# if not completed and not a ghost
1159
if parent not in graph:
1161
graph[version] = parents
1162
return topo_sort(graph.items())
1164
def get_ancestry_with_ghosts(self, versions):
1165
"""See VersionedFile.get_ancestry_with_ghosts."""
1166
# get a graph of all the mentioned versions:
1168
pending = set(versions)
1170
version = pending.pop()
1871
# Parse what we need to up front, this potentially trades off I/O
1872
# locality (.kndx and .knit in the same block group for the same file
1873
# id) for less checking in inner loops.
1874
prefixes = set(key[:-1] for key in keys)
1875
self._load_prefixes(prefixes)
1172
parents = self._cache[version][4]
1880
suffix_parents = self._kndx_cache[prefix][0][key[-1]][4]
1173
1881
except KeyError:
1178
# got the parents ok
1179
for parent in parents:
1180
if parent not in graph:
1182
graph[version] = parents
1183
return topo_sort(graph.items())
1185
def num_versions(self):
1186
return len(self._history)
1188
__len__ = num_versions
1190
def get_versions(self):
1191
return self._history
1193
def idx_to_name(self, idx):
1194
return self._history[idx]
1196
def lookup(self, version_id):
1197
assert version_id in self._cache
1198
return self._cache[version_id][5]
1200
def _version_list_to_index(self, versions):
1884
result[key] = tuple(prefix + (suffix,) for
1885
suffix in suffix_parents)
1888
def get_position(self, key):
1889
"""Return details needed to access the version.
1891
:return: a tuple (key, data position, size) to hand to the access
1892
logic to get the record.
1894
prefix, suffix = self._split_key(key)
1895
self._load_prefixes([prefix])
1896
entry = self._kndx_cache[prefix][0][suffix]
1897
return key, entry[2], entry[3]
1899
def _init_index(self, path, extra_lines=[]):
1900
"""Initialize an index."""
1902
sio.write(self.HEADER)
1903
sio.writelines(extra_lines)
1905
self._transport.put_file_non_atomic(path, sio,
1906
create_parent_dir=True)
1907
# self._create_parent_dir)
1908
# mode=self._file_mode,
1909
# dir_mode=self._dir_mode)
1912
"""Get all the keys in the collection.
1914
The keys are not ordered.
1917
# Identify all key prefixes.
1918
# XXX: A bit hacky, needs polish.
1919
if type(self._mapper) == ConstantMapper:
1923
for quoted_relpath in self._transport.iter_files_recursive():
1924
path, ext = os.path.splitext(quoted_relpath)
1926
prefixes = [self._mapper.unmap(path) for path in relpaths]
1927
self._load_prefixes(prefixes)
1928
for prefix in prefixes:
1929
for suffix in self._kndx_cache[prefix][1]:
1930
result.add(prefix + (suffix,))
1933
def _load_prefixes(self, prefixes):
1934
"""Load the indices for prefixes."""
1936
for prefix in prefixes:
1937
if prefix not in self._kndx_cache:
1938
# the load_data interface writes to these variables.
1941
self._filename = prefix
1943
path = self._mapper.map(prefix) + '.kndx'
1944
fp = self._transport.get(path)
1946
# _load_data may raise NoSuchFile if the target knit is
1948
_load_data(self, fp)
1951
self._kndx_cache[prefix] = (self._cache, self._history)
1956
self._kndx_cache[prefix] = ({}, [])
1957
if type(self._mapper) == ConstantMapper:
1958
# preserve behaviour for revisions.kndx etc.
1959
self._init_index(path)
1964
def _partition_keys(self, keys):
1965
"""Turn keys into a dict of prefix:suffix_list."""
1968
prefix_keys = result.setdefault(key[:-1], [])
1969
prefix_keys.append(key[-1])
1972
def _dictionary_compress(self, keys):
1973
"""Dictionary compress keys.
1975
:param keys: The keys to generate references to.
1976
:return: A string representation of keys. keys which are present are
1977
dictionary compressed, and others are emitted as fulltext with a
1201
1982
result_list = []
1202
for version in versions:
1203
if version in self._cache:
1983
prefix = keys[0][:-1]
1984
cache = self._kndx_cache[prefix][0]
1986
if key[:-1] != prefix:
1987
# kndx indices cannot refer across partitioned storage.
1988
raise ValueError("mismatched prefixes for %r" % keys)
1989
if key[-1] in cache:
1204
1990
# -- inlined lookup() --
1205
result_list.append(str(self._cache[version][5]))
1991
result_list.append(str(cache[key[-1]][5]))
1206
1992
# -- end lookup () --
1208
result_list.append('.' + version.encode('utf-8'))
1994
result_list.append('.' + key[-1])
1209
1995
return ' '.join(result_list)
1211
def add_version(self, version_id, options, pos, size, parents):
1212
"""Add a version record to the index."""
1213
self.add_versions(((version_id, options, pos, size, parents),))
1215
def add_versions(self, versions):
1216
"""Add multiple versions to the index.
1218
:param versions: a list of tuples:
1219
(version_id, options, pos, size, parents).
1222
for version_id, options, pos, size, parents in versions:
1223
line = "\n%s %s %s %s %s :" % (version_id.encode('utf-8'),
1227
self._version_list_to_index(parents))
1228
assert isinstance(line, str), \
1229
'content must be utf-8 encoded: %r' % (line,)
1231
self._transport.append(self._filename, StringIO(''.join(lines)))
1232
# cache after writing, so that a failed write leads to missing cache
1233
# entries not extra ones. XXX TODO: RBC 20060502 in the event of a
1234
# failure, reload the index or flush it or some such, to prevent
1235
# writing records that did complete twice.
1236
for version_id, options, pos, size, parents in versions:
1237
self._cache_version(version_id, options, pos, size, parents)
1239
def has_version(self, version_id):
1240
"""True if the version is in the index."""
1241
return self._cache.has_key(version_id)
1243
def get_position(self, version_id):
1244
"""Return data position and size of specified version."""
1245
return (self._cache[version_id][2], \
1246
self._cache[version_id][3])
1248
def get_method(self, version_id):
1249
"""Return compression method of specified version."""
1250
options = self._cache[version_id][1]
1251
if 'fulltext' in options:
1997
def _reset_cache(self):
1998
# Possibly this should be a LRU cache. A dictionary from key_prefix to
1999
# (cache_dict, history_vector) for parsed kndx files.
2000
self._kndx_cache = {}
2001
self._scope = self._get_scope()
2002
allow_writes = self._allow_writes()
2008
def _split_key(self, key):
2009
"""Split key into a prefix and suffix."""
2010
return key[:-1], key[-1]
2013
class _KnitGraphIndex(object):
2014
"""A KnitVersionedFiles index layered on GraphIndex."""
2016
def __init__(self, graph_index, is_locked, deltas=False, parents=True,
2018
"""Construct a KnitGraphIndex on a graph_index.
2020
:param graph_index: An implementation of bzrlib.index.GraphIndex.
2021
:param is_locked: A callback to check whether the object should answer
2023
:param deltas: Allow delta-compressed records.
2024
:param parents: If True, record knits parents, if not do not record
2026
:param add_callback: If not None, allow additions to the index and call
2027
this callback with a list of added GraphIndex nodes:
2028
[(node, value, node_refs), ...]
2029
:param is_locked: A callback, returns True if the index is locked and
2032
self._add_callback = add_callback
2033
self._graph_index = graph_index
2034
self._deltas = deltas
2035
self._parents = parents
2036
if deltas and not parents:
2037
# XXX: TODO: Delta tree and parent graph should be conceptually
2039
raise KnitCorrupt(self, "Cannot do delta compression without "
2041
self.has_graph = parents
2042
self._is_locked = is_locked
2045
return "%s(%r)" % (self.__class__.__name__, self._graph_index)
2047
def add_records(self, records, random_id=False):
2048
"""Add multiple records to the index.
2050
This function does not insert data into the Immutable GraphIndex
2051
backing the KnitGraphIndex, instead it prepares data for insertion by
2052
the caller and checks that it is safe to insert then calls
2053
self._add_callback with the prepared GraphIndex nodes.
2055
:param records: a list of tuples:
2056
(key, options, access_memo, parents).
2057
:param random_id: If True the ids being added were randomly generated
2058
and no check for existence will be performed.
2060
if not self._add_callback:
2061
raise errors.ReadOnlyError(self)
2062
# we hope there are no repositories with inconsistent parentage
2066
for (key, options, access_memo, parents) in records:
2068
parents = tuple(parents)
2069
index, pos, size = access_memo
2070
if 'no-eol' in options:
2074
value += "%d %d" % (pos, size)
2075
if not self._deltas:
2076
if 'line-delta' in options:
2077
raise KnitCorrupt(self, "attempt to add line-delta in non-delta knit")
2080
if 'line-delta' in options:
2081
node_refs = (parents, (parents[0],))
2083
node_refs = (parents, ())
2085
node_refs = (parents, )
2088
raise KnitCorrupt(self, "attempt to add node with parents "
2089
"in parentless index.")
2091
keys[key] = (value, node_refs)
2094
present_nodes = self._get_entries(keys)
2095
for (index, key, value, node_refs) in present_nodes:
2096
if (value[0] != keys[key][0][0] or
2097
node_refs != keys[key][1]):
2098
raise KnitCorrupt(self, "inconsistent details in add_records"
2099
": %s %s" % ((value, node_refs), keys[key]))
2103
for key, (value, node_refs) in keys.iteritems():
2104
result.append((key, value, node_refs))
2106
for key, (value, node_refs) in keys.iteritems():
2107
result.append((key, value))
2108
self._add_callback(result)
2110
def _check_read(self):
2111
"""raise if reads are not permitted."""
2112
if not self._is_locked():
2113
raise errors.ObjectNotLocked(self)
2115
def _check_write_ok(self):
2116
"""Assert if writes are not permitted."""
2117
if not self._is_locked():
2118
raise errors.ObjectNotLocked(self)
2120
def _compression_parent(self, an_entry):
2121
# return the key that an_entry is compressed against, or None
2122
# Grab the second parent list (as deltas implies parents currently)
2123
compression_parents = an_entry[3][1]
2124
if not compression_parents:
2126
if len(compression_parents) != 1:
2127
raise AssertionError(
2128
"Too many compression parents: %r" % compression_parents)
2129
return compression_parents[0]
2131
def get_build_details(self, keys):
2132
"""Get the method, index_memo and compression parent for version_ids.
2134
Ghosts are omitted from the result.
2136
:param keys: An iterable of keys.
2137
:return: A dict of key:
2138
(index_memo, compression_parent, parents, record_details).
2140
opaque structure to pass to read_records to extract the raw
2143
Content that this record is built upon, may be None
2145
Logical parents of this node
2147
extra information about the content which needs to be passed to
2148
Factory.parse_record
2152
entries = self._get_entries(keys, False)
2153
for entry in entries:
2155
if not self._parents:
2158
parents = entry[3][0]
2159
if not self._deltas:
2160
compression_parent_key = None
2162
compression_parent_key = self._compression_parent(entry)
2163
noeol = (entry[2][0] == 'N')
2164
if compression_parent_key:
2165
method = 'line-delta'
2168
result[key] = (self._node_to_position(entry),
2169
compression_parent_key, parents,
2173
def _get_entries(self, keys, check_present=False):
2174
"""Get the entries for keys.
2176
:param keys: An iterable of index key tuples.
2181
for node in self._graph_index.iter_entries(keys):
2183
found_keys.add(node[1])
2185
# adapt parentless index to the rest of the code.
2186
for node in self._graph_index.iter_entries(keys):
2187
yield node[0], node[1], node[2], ()
2188
found_keys.add(node[1])
2190
missing_keys = keys.difference(found_keys)
2192
raise RevisionNotPresent(missing_keys.pop(), self)
2194
def get_method(self, key):
2195
"""Return compression method of specified key."""
2196
return self._get_method(self._get_node(key))
2198
def _get_method(self, node):
2199
if not self._deltas:
1252
2200
return 'fulltext'
1254
assert 'line-delta' in options
2201
if self._compression_parent(node):
1255
2202
return 'line-delta'
1257
def get_options(self, version_id):
1258
return self._cache[version_id][1]
1260
def get_parents(self, version_id):
1261
"""Return parents of specified version ignoring ghosts."""
1262
return [parent for parent in self._cache[version_id][4]
1263
if parent in self._cache]
1265
def get_parents_with_ghosts(self, version_id):
1266
"""Return parents of specified version with ghosts."""
1267
return self._cache[version_id][4]
1269
def check_versions_present(self, version_ids):
1270
"""Check that all specified versions are present."""
1271
version_ids = set(version_ids)
1272
for version_id in list(version_ids):
1273
if version_id in self._cache:
1274
version_ids.remove(version_id)
1276
raise RevisionNotPresent(list(version_ids)[0], self.filename)
1279
class _KnitData(_KnitComponentFile):
1280
"""Contents of the knit data file"""
1282
HEADER = "# bzr knit data 8\n"
1284
def __init__(self, transport, filename, mode, create=False, file_mode=None):
1285
_KnitComponentFile.__init__(self, transport, filename, mode)
1286
self._checked = False
1288
self._transport.put(self._filename, StringIO(''), mode=file_mode)
1290
def clear_cache(self):
1291
"""Clear the record cache."""
1294
def _open_file(self):
1296
return self._transport.get(self._filename)
1301
def _record_to_data(self, version_id, digest, lines):
1302
"""Convert version_id, digest, lines into a raw data block.
1304
:return: (len, a StringIO instance with the raw data ready to read.)
1307
data_file = GzipFile(None, mode='wb', fileobj=sio)
1308
data_file.writelines(chain(
1309
["version %s %d %s\n" % (version_id.encode('utf-8'),
1313
["end %s\n" % version_id.encode('utf-8')]))
1320
def add_raw_record(self, raw_data):
1321
"""Append a prepared record to the data file.
1323
:return: the offset in the data file raw_data was written.
1325
assert isinstance(raw_data, str), 'data must be plain bytes'
1326
return self._transport.append(self._filename, StringIO(raw_data))
1328
def add_record(self, version_id, digest, lines):
1329
"""Write new text record to disk. Returns the position in the
1330
file where it was written."""
1331
size, sio = self._record_to_data(version_id, digest, lines)
1333
start_pos = self._transport.append(self._filename, sio)
1334
return start_pos, size
1336
def _parse_record_header(self, version_id, raw_data):
1337
"""Parse a record header for consistency.
1339
:return: the header and the decompressor stream.
1340
as (stream, header_record)
1342
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
1343
rec = df.readline().split()
1345
raise KnitCorrupt(self._filename, 'unexpected number of elements in record header')
1346
if rec[1].decode('utf-8')!= version_id:
1347
raise KnitCorrupt(self._filename,
1348
'unexpected version, wanted %r, got %r' % (
1349
version_id, rec[1]))
1352
def _parse_record(self, version_id, data):
1354
# 4168 calls in 2880 217 internal
1355
# 4168 calls to _parse_record_header in 2121
1356
# 4168 calls to readlines in 330
1357
df, rec = self._parse_record_header(version_id, data)
1358
record_contents = df.readlines()
1359
l = record_contents.pop()
1360
assert len(record_contents) == int(rec[2])
1361
if l.decode('utf-8') != 'end %s\n' % version_id:
1362
raise KnitCorrupt(self._filename, 'unexpected version end line %r, wanted %r'
1365
return record_contents, rec[3]
1367
def read_records_iter_raw(self, records):
1368
"""Read text records from data file and yield raw data.
1370
This unpacks enough of the text record to validate the id is
1371
as expected but thats all.
1373
It will actively recompress currently cached records on the
1374
basis that that is cheaper than I/O activity.
1376
# setup an iterator of the external records:
1377
# uses readv so nice and fast we hope.
1379
# grab the disk data needed.
1380
raw_records = self._transport.readv(self._filename,
1381
[(pos, size) for version_id, pos, size in records])
1383
for version_id, pos, size in records:
1384
pos, data = raw_records.next()
1385
# validate the header
1386
df, rec = self._parse_record_header(version_id, data)
1388
yield version_id, data
1390
def read_records_iter(self, records):
1391
"""Read text records from data file and yield result.
1393
Each passed record is a tuple of (version_id, pos, len) and
1394
will be read in the given order. Yields (version_id,
1397
if len(records) == 0:
1400
# 60890 calls for 4168 extractions in 5045, 683 internal.
1401
# 4168 calls to readv in 1411
1402
# 4168 calls to parse_record in 2880
1404
# Get unique records, sorted by position
1405
needed_records = sorted(set(records), key=operator.itemgetter(1))
1407
# We take it that the transport optimizes the fetching as good
1408
# as possible (ie, reads continuous ranges.)
1409
response = self._transport.readv(self._filename,
1410
[(pos, size) for version_id, pos, size in needed_records])
1413
for (record_id, pos, size), (pos, data) in \
1414
izip(iter(needed_records), response):
1415
content, digest = self._parse_record(record_id, data)
1416
record_map[record_id] = (digest, content)
1418
for version_id, pos, size in records:
1419
digest, content = record_map[version_id]
1420
yield version_id, content, digest
1422
def read_records(self, records):
1423
"""Read records into a dictionary."""
1425
for record_id, content, digest in self.read_records_iter(records):
1426
components[record_id] = (content, digest)
1430
class InterKnit(InterVersionedFile):
1431
"""Optimised code paths for knit to knit operations."""
1433
_matching_file_from_factory = KnitVersionedFile
1434
_matching_file_to_factory = KnitVersionedFile
1437
def is_compatible(source, target):
1438
"""Be compatible with knits. """
1440
return (isinstance(source, KnitVersionedFile) and
1441
isinstance(target, KnitVersionedFile))
1442
except AttributeError:
1445
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
1446
"""See InterVersionedFile.join."""
1447
assert isinstance(self.source, KnitVersionedFile)
1448
assert isinstance(self.target, KnitVersionedFile)
1450
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
1455
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1457
version_ids = list(version_ids)
1458
if None in version_ids:
1459
version_ids.remove(None)
1461
self.source_ancestry = set(self.source.get_ancestry(version_ids))
1462
this_versions = set(self.target._index.get_versions())
1463
needed_versions = self.source_ancestry - this_versions
1464
cross_check_versions = self.source_ancestry.intersection(this_versions)
1465
mismatched_versions = set()
1466
for version in cross_check_versions:
1467
# scan to include needed parents.
1468
n1 = set(self.target.get_parents_with_ghosts(version))
1469
n2 = set(self.source.get_parents_with_ghosts(version))
1471
# FIXME TEST this check for cycles being introduced works
1472
# the logic is we have a cycle if in our graph we are an
1473
# ancestor of any of the n2 revisions.
1479
parent_ancestors = self.source.get_ancestry(parent)
1480
if version in parent_ancestors:
1481
raise errors.GraphCycleError([parent, version])
1482
# ensure this parent will be available later.
1483
new_parents = n2.difference(n1)
1484
needed_versions.update(new_parents.difference(this_versions))
1485
mismatched_versions.add(version)
1487
if not needed_versions and not mismatched_versions:
1489
full_list = topo_sort(self.source.get_graph())
1491
version_list = [i for i in full_list if (not self.target.has_version(i)
1492
and i in needed_versions)]
1496
copy_queue_records = []
1498
for version_id in version_list:
1499
options = self.source._index.get_options(version_id)
1500
parents = self.source._index.get_parents_with_ghosts(version_id)
1501
# check that its will be a consistent copy:
1502
for parent in parents:
1503
# if source has the parent, we must :
1504
# * already have it or
1505
# * have it scheduled already
1506
# otherwise we don't care
1507
assert (self.target.has_version(parent) or
1508
parent in copy_set or
1509
not self.source.has_version(parent))
1510
data_pos, data_size = self.source._index.get_position(version_id)
1511
copy_queue_records.append((version_id, data_pos, data_size))
1512
copy_queue.append((version_id, options, parents))
1513
copy_set.add(version_id)
1515
# data suck the join:
1517
total = len(version_list)
1520
for (version_id, raw_data), \
1521
(version_id2, options, parents) in \
1522
izip(self.source._data.read_records_iter_raw(copy_queue_records),
1524
assert version_id == version_id2, 'logic error, inconsistent results'
1526
pb.update("Joining knit", count, total)
1527
raw_records.append((version_id, options, parents, len(raw_data)))
1528
raw_datum.append(raw_data)
1529
self.target._add_raw_records(raw_records, ''.join(raw_datum))
1531
for version in mismatched_versions:
1532
# FIXME RBC 20060309 is this needed?
1533
n1 = set(self.target.get_parents_with_ghosts(version))
1534
n2 = set(self.source.get_parents_with_ghosts(version))
1535
# write a combined record to our history preserving the current
1536
# parents as first in the list
1537
new_parents = self.target.get_parents_with_ghosts(version) + list(n2.difference(n1))
1538
self.target.fix_parents(version, new_parents)
1544
InterVersionedFile.register_optimiser(InterKnit)
1547
class WeaveToKnit(InterVersionedFile):
1548
"""Optimised code paths for weave to knit operations."""
1550
_matching_file_from_factory = bzrlib.weave.WeaveFile
1551
_matching_file_to_factory = KnitVersionedFile
1554
def is_compatible(source, target):
1555
"""Be compatible with weaves to knits."""
1557
return (isinstance(source, bzrlib.weave.Weave) and
1558
isinstance(target, KnitVersionedFile))
1559
except AttributeError:
1562
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
1563
"""See InterVersionedFile.join."""
1564
assert isinstance(self.source, bzrlib.weave.Weave)
1565
assert isinstance(self.target, KnitVersionedFile)
1567
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
1572
pb = bzrlib.ui.ui_factory.nested_progress_bar()
1574
version_ids = list(version_ids)
1576
self.source_ancestry = set(self.source.get_ancestry(version_ids))
1577
this_versions = set(self.target._index.get_versions())
1578
needed_versions = self.source_ancestry - this_versions
1579
cross_check_versions = self.source_ancestry.intersection(this_versions)
1580
mismatched_versions = set()
1581
for version in cross_check_versions:
1582
# scan to include needed parents.
1583
n1 = set(self.target.get_parents_with_ghosts(version))
1584
n2 = set(self.source.get_parents(version))
1585
# if all of n2's parents are in n1, then its fine.
1586
if n2.difference(n1):
1587
# FIXME TEST this check for cycles being introduced works
1588
# the logic is we have a cycle if in our graph we are an
1589
# ancestor of any of the n2 revisions.
1595
parent_ancestors = self.source.get_ancestry(parent)
1596
if version in parent_ancestors:
1597
raise errors.GraphCycleError([parent, version])
1598
# ensure this parent will be available later.
1599
new_parents = n2.difference(n1)
1600
needed_versions.update(new_parents.difference(this_versions))
1601
mismatched_versions.add(version)
1603
if not needed_versions and not mismatched_versions:
1605
full_list = topo_sort(self.source.get_graph())
1607
version_list = [i for i in full_list if (not self.target.has_version(i)
1608
and i in needed_versions)]
1612
total = len(version_list)
1613
for version_id in version_list:
1614
pb.update("Converting to knit", count, total)
1615
parents = self.source.get_parents(version_id)
1616
# check that its will be a consistent copy:
1617
for parent in parents:
1618
# if source has the parent, we must already have it
1619
assert (self.target.has_version(parent))
1620
self.target.add_lines(
1621
version_id, parents, self.source.get_lines(version_id))
1624
for version in mismatched_versions:
1625
# FIXME RBC 20060309 is this needed?
1626
n1 = set(self.target.get_parents_with_ghosts(version))
1627
n2 = set(self.source.get_parents(version))
1628
# write a combined record to our history preserving the current
1629
# parents as first in the list
1630
new_parents = self.target.get_parents_with_ghosts(version) + list(n2.difference(n1))
1631
self.target.fix_parents(version, new_parents)
1637
InterVersionedFile.register_optimiser(WeaveToKnit)
1640
class KnitSequenceMatcher(difflib.SequenceMatcher):
1641
"""Knit tuned sequence matcher.
1643
This is based on profiling of difflib which indicated some improvements
1644
for our usage pattern.
2206
def _get_node(self, key):
2208
return list(self._get_entries([key]))[0]
2210
raise RevisionNotPresent(key, self)
2212
def get_options(self, key):
2213
"""Return a list representing options.
2217
node = self._get_node(key)
2218
options = [self._get_method(node)]
2219
if node[2][0] == 'N':
2220
options.append('no-eol')
2223
def get_parent_map(self, keys):
2224
"""Get a map of the parents of keys.
2226
:param keys: The keys to look up parents for.
2227
:return: A mapping from keys to parents. Absent keys are absent from
2231
nodes = self._get_entries(keys)
2235
result[node[1]] = node[3][0]
2238
result[node[1]] = None
2241
def get_position(self, key):
2242
"""Return details needed to access the version.
2244
:return: a tuple (index, data position, size) to hand to the access
2245
logic to get the record.
2247
node = self._get_node(key)
2248
return self._node_to_position(node)
2251
"""Get all the keys in the collection.
2253
The keys are not ordered.
2256
return [node[1] for node in self._graph_index.iter_all_entries()]
2258
def _node_to_position(self, node):
2259
"""Convert an index value to position details."""
2260
bits = node[2][1:].split(' ')
2261
return node[0], int(bits[0]), int(bits[1])
2264
class _KnitKeyAccess(object):
2265
"""Access to records in .knit files."""
2267
def __init__(self, transport, mapper):
2268
"""Create a _KnitKeyAccess with transport and mapper.
2270
:param transport: The transport the access object is rooted at.
2271
:param mapper: The mapper used to map keys to .knit files.
2273
self._transport = transport
2274
self._mapper = mapper
2276
def add_raw_records(self, key_sizes, raw_data):
2277
"""Add raw knit bytes to a storage area.
2279
The data is spooled to the container writer in one bytes-record per
2282
:param sizes: An iterable of tuples containing the key and size of each
2284
:param raw_data: A bytestring containing the data.
2285
:return: A list of memos to retrieve the record later. Each memo is an
2286
opaque index memo. For _KnitKeyAccess the memo is (key, pos,
2287
length), where the key is the record key.
2289
if type(raw_data) != str:
2290
raise AssertionError(
2291
'data must be plain bytes was %s' % type(raw_data))
2294
# TODO: This can be tuned for writing to sftp and other servers where
2295
# append() is relatively expensive by grouping the writes to each key
2297
for key, size in key_sizes:
2298
path = self._mapper.map(key)
2300
base = self._transport.append_bytes(path + '.knit',
2301
raw_data[offset:offset+size])
2302
except errors.NoSuchFile:
2303
self._transport.mkdir(osutils.dirname(path))
2304
base = self._transport.append_bytes(path + '.knit',
2305
raw_data[offset:offset+size])
2309
result.append((key, base, size))
2312
def get_raw_records(self, memos_for_retrieval):
2313
"""Get the raw bytes for a records.
2315
:param memos_for_retrieval: An iterable containing the access memo for
2316
retrieving the bytes.
2317
:return: An iterator over the bytes of the records.
2319
# first pass, group into same-index request to minimise readv's issued.
2321
current_prefix = None
2322
for (key, offset, length) in memos_for_retrieval:
2323
if current_prefix == key[:-1]:
2324
current_list.append((offset, length))
2326
if current_prefix is not None:
2327
request_lists.append((current_prefix, current_list))
2328
current_prefix = key[:-1]
2329
current_list = [(offset, length)]
2330
# handle the last entry
2331
if current_prefix is not None:
2332
request_lists.append((current_prefix, current_list))
2333
for prefix, read_vector in request_lists:
2334
path = self._mapper.map(prefix) + '.knit'
2335
for pos, data in self._transport.readv(path, read_vector):
2339
class _DirectPackAccess(object):
2340
"""Access to data in one or more packs with less translation."""
2342
def __init__(self, index_to_packs):
2343
"""Create a _DirectPackAccess object.
2345
:param index_to_packs: A dict mapping index objects to the transport
2346
and file names for obtaining data.
2348
self._container_writer = None
2349
self._write_index = None
2350
self._indices = index_to_packs
2352
def add_raw_records(self, key_sizes, raw_data):
2353
"""Add raw knit bytes to a storage area.
2355
The data is spooled to the container writer in one bytes-record per
2358
:param sizes: An iterable of tuples containing the key and size of each
2360
:param raw_data: A bytestring containing the data.
2361
:return: A list of memos to retrieve the record later. Each memo is an
2362
opaque index memo. For _DirectPackAccess the memo is (index, pos,
2363
length), where the index field is the write_index object supplied
2364
to the PackAccess object.
2366
if type(raw_data) != str:
2367
raise AssertionError(
2368
'data must be plain bytes was %s' % type(raw_data))
2371
for key, size in key_sizes:
2372
p_offset, p_length = self._container_writer.add_bytes_record(
2373
raw_data[offset:offset+size], [])
2375
result.append((self._write_index, p_offset, p_length))
2378
def get_raw_records(self, memos_for_retrieval):
2379
"""Get the raw bytes for a records.
2381
:param memos_for_retrieval: An iterable containing the (index, pos,
2382
length) memo for retrieving the bytes. The Pack access method
2383
looks up the pack to use for a given record in its index_to_pack
2385
:return: An iterator over the bytes of the records.
2387
# first pass, group into same-index requests
2389
current_index = None
2390
for (index, offset, length) in memos_for_retrieval:
2391
if current_index == index:
2392
current_list.append((offset, length))
2394
if current_index is not None:
2395
request_lists.append((current_index, current_list))
2396
current_index = index
2397
current_list = [(offset, length)]
2398
# handle the last entry
2399
if current_index is not None:
2400
request_lists.append((current_index, current_list))
2401
for index, offsets in request_lists:
2402
transport, path = self._indices[index]
2403
reader = pack.make_readv_reader(transport, path, offsets)
2404
for names, read_func in reader.iter_records():
2405
yield read_func(None)
2407
def set_writer(self, writer, index, transport_packname):
2408
"""Set a writer to use for adding data."""
2409
if index is not None:
2410
self._indices[index] = transport_packname
2411
self._container_writer = writer
2412
self._write_index = index
2415
# Deprecated, use PatienceSequenceMatcher instead
2416
KnitSequenceMatcher = patiencediff.PatienceSequenceMatcher
2419
def annotate_knit(knit, revision_id):
2420
"""Annotate a knit with no cached annotations.
2422
This implementation is for knits with no cached annotations.
2423
It will work for knits with cached annotations, but this is not
1647
def find_longest_match(self, alo, ahi, blo, bhi):
1648
"""Find longest matching block in a[alo:ahi] and b[blo:bhi].
1650
If isjunk is not defined:
1652
Return (i,j,k) such that a[i:i+k] is equal to b[j:j+k], where
1653
alo <= i <= i+k <= ahi
1654
blo <= j <= j+k <= bhi
1655
and for all (i',j',k') meeting those conditions,
1658
and if i == i', j <= j'
1660
In other words, of all maximal matching blocks, return one that
1661
starts earliest in a, and of all those maximal matching blocks that
1662
start earliest in a, return the one that starts earliest in b.
1664
>>> s = SequenceMatcher(None, " abcd", "abcd abcd")
1665
>>> s.find_longest_match(0, 5, 0, 9)
1668
If isjunk is defined, first the longest matching block is
1669
determined as above, but with the additional restriction that no
1670
junk element appears in the block. Then that block is extended as
1671
far as possible by matching (only) junk elements on both sides. So
1672
the resulting block never matches on junk except as identical junk
1673
happens to be adjacent to an "interesting" match.
1675
Here's the same example as before, but considering blanks to be
1676
junk. That prevents " abcd" from matching the " abcd" at the tail
1677
end of the second sequence directly. Instead only the "abcd" can
1678
match, and matches the leftmost "abcd" in the second sequence:
1680
>>> s = SequenceMatcher(lambda x: x==" ", " abcd", "abcd abcd")
1681
>>> s.find_longest_match(0, 5, 0, 9)
1684
If no blocks match, return (alo, blo, 0).
1686
>>> s = SequenceMatcher(None, "ab", "c")
1687
>>> s.find_longest_match(0, 2, 0, 1)
1691
# CAUTION: stripping common prefix or suffix would be incorrect.
1695
# Longest matching block is "ab", but if common prefix is
1696
# stripped, it's "a" (tied with "b"). UNIX(tm) diff does so
1697
# strip, so ends up claiming that ab is changed to acab by
1698
# inserting "ca" in the middle. That's minimal but unintuitive:
1699
# "it's obvious" that someone inserted "ac" at the front.
1700
# Windiff ends up at the same place as diff, but by pairing up
1701
# the unique 'b's and then matching the first two 'a's.
1703
a, b, b2j, isbjunk = self.a, self.b, self.b2j, self.isbjunk
1704
besti, bestj, bestsize = alo, blo, 0
1705
# find longest junk-free match
1706
# during an iteration of the loop, j2len[j] = length of longest
1707
# junk-free match ending with a[i-1] and b[j]
1711
for i in xrange(alo, ahi):
1712
# look at all instances of a[i] in b; note that because
1713
# b2j has no junk keys, the loop is skipped if a[i] is junk
1714
j2lenget = j2len.get
1717
# changing b2j.get(a[i], nothing) to a try:KeyError pair produced the
1718
# following improvement
1719
# 704 0 4650.5320 2620.7410 bzrlib.knit:1336(find_longest_match)
1720
# +326674 0 1655.1210 1655.1210 +<method 'get' of 'dict' objects>
1721
# +76519 0 374.6700 374.6700 +<method 'has_key' of 'dict' objects>
1723
# 704 0 3733.2820 2209.6520 bzrlib.knit:1336(find_longest_match)
1724
# +211400 0 1147.3520 1147.3520 +<method 'get' of 'dict' objects>
1725
# +76519 0 376.2780 376.2780 +<method 'has_key' of 'dict' objects>
2426
annotator = _KnitAnnotator(knit)
2427
return iter(annotator.annotate(revision_id))
2430
class _KnitAnnotator(object):
2431
"""Build up the annotations for a text."""
2433
def __init__(self, knit):
2436
# Content objects, differs from fulltexts because of how final newlines
2437
# are treated by knits. the content objects here will always have a
2439
self._fulltext_contents = {}
2441
# Annotated lines of specific revisions
2442
self._annotated_lines = {}
2444
# Track the raw data for nodes that we could not process yet.
2445
# This maps the revision_id of the base to a list of children that will
2446
# annotated from it.
2447
self._pending_children = {}
2449
# Nodes which cannot be extracted
2450
self._ghosts = set()
2452
# Track how many children this node has, so we know if we need to keep
2454
self._annotate_children = {}
2455
self._compression_children = {}
2457
self._all_build_details = {}
2458
# The children => parent revision_id graph
2459
self._revision_id_graph = {}
2461
self._heads_provider = None
2463
self._nodes_to_keep_annotations = set()
2464
self._generations_until_keep = 100
2466
def set_generations_until_keep(self, value):
2467
"""Set the number of generations before caching a node.
2469
Setting this to -1 will cache every merge node, setting this higher
2470
will cache fewer nodes.
2472
self._generations_until_keep = value
2474
def _add_fulltext_content(self, revision_id, content_obj):
2475
self._fulltext_contents[revision_id] = content_obj
2476
# TODO: jam 20080305 It might be good to check the sha1digest here
2477
return content_obj.text()
2479
def _check_parents(self, child, nodes_to_annotate):
2480
"""Check if all parents have been processed.
2482
:param child: A tuple of (rev_id, parents, raw_content)
2483
:param nodes_to_annotate: If child is ready, add it to
2484
nodes_to_annotate, otherwise put it back in self._pending_children
2486
for parent_id in child[1]:
2487
if (parent_id not in self._annotated_lines):
2488
# This parent is present, but another parent is missing
2489
self._pending_children.setdefault(parent_id,
2493
# This one is ready to be processed
2494
nodes_to_annotate.append(child)
2496
def _add_annotation(self, revision_id, fulltext, parent_ids,
2497
left_matching_blocks=None):
2498
"""Add an annotation entry.
2500
All parents should already have been annotated.
2501
:return: A list of children that now have their parents satisfied.
2503
a = self._annotated_lines
2504
annotated_parent_lines = [a[p] for p in parent_ids]
2505
annotated_lines = list(annotate.reannotate(annotated_parent_lines,
2506
fulltext, revision_id, left_matching_blocks,
2507
heads_provider=self._get_heads_provider()))
2508
self._annotated_lines[revision_id] = annotated_lines
2509
for p in parent_ids:
2510
ann_children = self._annotate_children[p]
2511
ann_children.remove(revision_id)
2512
if (not ann_children
2513
and p not in self._nodes_to_keep_annotations):
2514
del self._annotated_lines[p]
2515
del self._all_build_details[p]
2516
if p in self._fulltext_contents:
2517
del self._fulltext_contents[p]
2518
# Now that we've added this one, see if there are any pending
2519
# deltas to be done, certainly this parent is finished
2520
nodes_to_annotate = []
2521
for child in self._pending_children.pop(revision_id, []):
2522
self._check_parents(child, nodes_to_annotate)
2523
return nodes_to_annotate
2525
def _get_build_graph(self, key):
2526
"""Get the graphs for building texts and annotations.
2528
The data you need for creating a full text may be different than the
2529
data you need to annotate that text. (At a minimum, you need both
2530
parents to create an annotation, but only need 1 parent to generate the
2533
:return: A list of (key, index_memo) records, suitable for
2534
passing to read_records_iter to start reading in the raw data fro/
2537
if key in self._annotated_lines:
2540
pending = set([key])
2545
# get all pending nodes
2547
this_iteration = pending
2548
build_details = self._knit._index.get_build_details(this_iteration)
2549
self._all_build_details.update(build_details)
2550
# new_nodes = self._knit._index._get_entries(this_iteration)
2552
for key, details in build_details.iteritems():
2553
(index_memo, compression_parent, parents,
2554
record_details) = details
2555
self._revision_id_graph[key] = parents
2556
records.append((key, index_memo))
2557
# Do we actually need to check _annotated_lines?
2558
pending.update(p for p in parents
2559
if p not in self._all_build_details)
2560
if compression_parent:
2561
self._compression_children.setdefault(compression_parent,
2564
for parent in parents:
2565
self._annotate_children.setdefault(parent,
2567
num_gens = generation - kept_generation
2568
if ((num_gens >= self._generations_until_keep)
2569
and len(parents) > 1):
2570
kept_generation = generation
2571
self._nodes_to_keep_annotations.add(key)
2573
missing_versions = this_iteration.difference(build_details.keys())
2574
self._ghosts.update(missing_versions)
2575
for missing_version in missing_versions:
2576
# add a key, no parents
2577
self._revision_id_graph[missing_version] = ()
2578
pending.discard(missing_version) # don't look for it
2579
if self._ghosts.intersection(self._compression_children):
2581
"We cannot have nodes which have a ghost compression parent:\n"
2583
"compression children: %r"
2584
% (self._ghosts, self._compression_children))
2585
# Cleanout anything that depends on a ghost so that we don't wait for
2586
# the ghost to show up
2587
for node in self._ghosts:
2588
if node in self._annotate_children:
2589
# We won't be building this node
2590
del self._annotate_children[node]
2591
# Generally we will want to read the records in reverse order, because
2592
# we find the parent nodes after the children
2596
def _annotate_records(self, records):
2597
"""Build the annotations for the listed records."""
2598
# We iterate in the order read, rather than a strict order requested
2599
# However, process what we can, and put off to the side things that
2600
# still need parents, cleaning them up when those parents are
2602
for (rev_id, record,
2603
digest) in self._knit._read_records_iter(records):
2604
if rev_id in self._annotated_lines:
2606
parent_ids = self._revision_id_graph[rev_id]
2607
parent_ids = [p for p in parent_ids if p not in self._ghosts]
2608
details = self._all_build_details[rev_id]
2609
(index_memo, compression_parent, parents,
2610
record_details) = details
2611
nodes_to_annotate = []
2612
# TODO: Remove the punning between compression parents, and
2613
# parent_ids, we should be able to do this without assuming
2615
if len(parent_ids) == 0:
2616
# There are no parents for this node, so just add it
2617
# TODO: This probably needs to be decoupled
2618
fulltext_content, delta = self._knit._factory.parse_record(
2619
rev_id, record, record_details, None)
2620
fulltext = self._add_fulltext_content(rev_id, fulltext_content)
2621
nodes_to_annotate.extend(self._add_annotation(rev_id, fulltext,
2622
parent_ids, left_matching_blocks=None))
2624
child = (rev_id, parent_ids, record)
2625
# Check if all the parents are present
2626
self._check_parents(child, nodes_to_annotate)
2627
while nodes_to_annotate:
2628
# Should we use a queue here instead of a stack?
2629
(rev_id, parent_ids, record) = nodes_to_annotate.pop()
2630
(index_memo, compression_parent, parents,
2631
record_details) = self._all_build_details[rev_id]
2632
if compression_parent is not None:
2633
comp_children = self._compression_children[compression_parent]
2634
if rev_id not in comp_children:
2635
raise AssertionError("%r not in compression children %r"
2636
% (rev_id, comp_children))
2637
# If there is only 1 child, it is safe to reuse this
2639
reuse_content = (len(comp_children) == 1
2640
and compression_parent not in
2641
self._nodes_to_keep_annotations)
2643
# Remove it from the cache since it will be changing
2644
parent_fulltext_content = self._fulltext_contents.pop(compression_parent)
2645
# Make sure to copy the fulltext since it might be
2647
parent_fulltext = list(parent_fulltext_content.text())
2649
parent_fulltext_content = self._fulltext_contents[compression_parent]
2650
parent_fulltext = parent_fulltext_content.text()
2651
comp_children.remove(rev_id)
2652
fulltext_content, delta = self._knit._factory.parse_record(
2653
rev_id, record, record_details,
2654
parent_fulltext_content,
2655
copy_base_content=(not reuse_content))
2656
fulltext = self._add_fulltext_content(rev_id,
2658
blocks = KnitContent.get_line_delta_blocks(delta,
2659
parent_fulltext, fulltext)
2661
fulltext_content = self._knit._factory.parse_fulltext(
2663
fulltext = self._add_fulltext_content(rev_id,
2666
nodes_to_annotate.extend(
2667
self._add_annotation(rev_id, fulltext, parent_ids,
2668
left_matching_blocks=blocks))
2670
def _get_heads_provider(self):
2671
"""Create a heads provider for resolving ancestry issues."""
2672
if self._heads_provider is not None:
2673
return self._heads_provider
2674
parent_provider = _mod_graph.DictParentsProvider(
2675
self._revision_id_graph)
2676
graph_obj = _mod_graph.Graph(parent_provider)
2677
head_cache = _mod_graph.FrozenHeadsCache(graph_obj)
2678
self._heads_provider = head_cache
2681
def annotate(self, key):
2682
"""Return the annotated fulltext at the given key.
2684
:param key: The key to annotate.
2686
if True or len(self._knit._fallback_vfs) > 0:
2687
# stacked knits can't use the fast path at present.
2688
return self._simple_annotate(key)
2689
records = self._get_build_graph(key)
2690
if key in self._ghosts:
2691
raise errors.RevisionNotPresent(key, self._knit)
2692
self._annotate_records(records)
2693
return self._annotated_lines[key]
2695
def _simple_annotate(self, key):
2696
"""Return annotated fulltext, rediffing from the full texts.
2698
This is slow but makes no assumptions about the repository
2699
being able to produce line deltas.
2701
# TODO: this code generates a parent maps of present ancestors; it
2702
# could be split out into a separate method, and probably should use
2703
# iter_ancestry instead. -- mbp and robertc 20080704
2704
graph = _mod_graph.Graph(self._knit)
2705
head_cache = _mod_graph.FrozenHeadsCache(graph)
2706
search = graph._make_breadth_first_searcher([key])
2710
present, ghosts = search.next_with_ghosts()
2711
except StopIteration:
2713
keys.update(present)
2714
parent_map = self._knit.get_parent_map(keys)
2716
reannotate = annotate.reannotate
2717
for record in self._knit.get_record_stream(keys, 'topological', True):
2719
fulltext = split_lines(record.get_bytes_as('fulltext'))
2720
parents = parent_map[key]
2721
if parents is not None:
2722
parent_lines = [parent_cache[parent] for parent in parent_map[key]]
1737
k = newj2len[j] = 1 + j2lenget(-1 + j, 0)
1739
besti, bestj, bestsize = 1 + i-k, 1 + j-k, k
1742
# Extend the best by non-junk elements on each end. In particular,
1743
# "popular" non-junk elements aren't in b2j, which greatly speeds
1744
# the inner loop above, but also means "the best" match so far
1745
# doesn't contain any junk *or* popular non-junk elements.
1746
while besti > alo and bestj > blo and \
1747
not isbjunk(b[bestj-1]) and \
1748
a[besti-1] == b[bestj-1]:
1749
besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
1750
while besti+bestsize < ahi and bestj+bestsize < bhi and \
1751
not isbjunk(b[bestj+bestsize]) and \
1752
a[besti+bestsize] == b[bestj+bestsize]:
1755
# Now that we have a wholly interesting match (albeit possibly
1756
# empty!), we may as well suck up the matching junk on each
1757
# side of it too. Can't think of a good reason not to, and it
1758
# saves post-processing the (possibly considerable) expense of
1759
# figuring out what to do with it. In the case of an empty
1760
# interesting match, this is clearly the right thing to do,
1761
# because no other kind of match is possible in the regions.
1762
while besti > alo and bestj > blo and \
1763
isbjunk(b[bestj-1]) and \
1764
a[besti-1] == b[bestj-1]:
1765
besti, bestj, bestsize = besti-1, bestj-1, bestsize+1
1766
while besti+bestsize < ahi and bestj+bestsize < bhi and \
1767
isbjunk(b[bestj+bestsize]) and \
1768
a[besti+bestsize] == b[bestj+bestsize]:
1769
bestsize = bestsize + 1
1771
return besti, bestj, bestsize
2725
parent_cache[key] = list(
2726
reannotate(parent_lines, fulltext, key, None, head_cache))
2728
return parent_cache[key]
2730
raise errors.RevisionNotPresent(key, self._knit)
2734
from bzrlib._knit_load_data_c import _load_data_c as _load_data
2736
from bzrlib._knit_load_data_py import _load_data_py as _load_data