« back to all changes in this revision
Viewing changes to bzrlib/knit.py
- browse files at revision 2900.2.24
- compare with another revision
- download diff
- download tarball
- view history from revision 2900.2.24
- Committer: Vincent Ladeuil
- Date: 2007-11-04 15:24:27 UTC
- mto: (2961.1.1 trunk)
- mto: This revision was merged to the branch mainline in revision 2962.
- Revision ID: v.ladeuil+lp@free.fr-20071104152427-p9k7e4toywa87wfc
Review feedback.
* doc/en/user-guide/authentication_conf.txt:
New file. Authentication configuration file documentation.
* doc/en/user-guide/configuration.txt:
Slight modifications, add authentication.conf reference.
* doc/en/mini-tutorial/index.txt:
Fix make docs warning.
* doc/developers/authentication-ring.txt:
Small cleanups noticed during
doc/en/user-guide/authentication_conf.txt redaction.
* bzrlib/transport/http/_urllib.py:
(HttpTransport_urllib._perform): Use a dict() instead of {} syntax.
* bzrlib/tests/blackbox/test_whoami.py:
(TestWhoami.test_whoami_branch): Delete BZREMAIL related tests.
* bzrlib/config.py:
(Config.username): BZREMAIL deleted, has been obsolete for more
than a year.
(AuthenticationConfig.__init__): Review feedback, since keeping a
callback as an attribute is useless, call it now and keep the
filename itself as an attribute.
(AuthenticationConfig.get_credentials): Use a dict() instead of {}
syntax.
* NEWS:
Updated as per Martin's suggestion.
* doc/en/user-guide/authentication_conf.txt:
New file. Authentication configuration file documentation.
* doc/en/user-guide/configuration.txt:
Slight modifications, add authentication.conf reference.
* doc/en/mini-tutorial/index.txt:
Fix make docs warning.
* doc/developers/authentication-ring.txt:
Small cleanups noticed during
doc/en/user-guide/authentication_conf.txt redaction.
* bzrlib/transport/http/_urllib.py:
(HttpTransport_urllib._perform): Use a dict() instead of {} syntax.
* bzrlib/tests/blackbox/test_whoami.py:
(TestWhoami.test_whoami_branch): Delete BZREMAIL related tests.
* bzrlib/config.py:
(Config.username): BZREMAIL deleted, has been obsolete for more
than a year.
(AuthenticationConfig.__init__): Review feedback, since keeping a
callback as an attribute is useless, call it now and keep the
filename itself as an attribute.
(AuthenticationConfig.get_credentials): Use a dict() instead of {}
syntax.
* NEWS:
Updated as per Martin's suggestion.
- files added:
- bzrlib/plugins/multiparent.py
- bzrlib/tests/interversionedfile_implementations
- bzrlib/tests/revisionstore_implementations
- files removed:
- bzrlib/_btree_serializer_c.pyx
- bzrlib/help_topics
- bzrlib/help_topics/en
- bzrlib/tests/per_repository_reference
- bzrlib/transport/ftp
- doc/en/admin-guide
- doc/en/tutorials
- doc/en/user-guide/images
- doc/es
- doc/es/guia-desarrollador
- doc/es/guia-usuario
- doc/es/mini-tutorial
- doc/es/notas-version
- doc/es/referencia
- doc/es/referencia-rapida
- tools/packaging
- files renamed:
- bzrlib/help_topics/__init__.py => bzrlib/help_topics.py
- bzrlib/plugins/launchpad/lp_directory.py => bzrlib/plugins/launchpad/lp_indirect.py
- bzrlib/plugins/launchpad/test_lp_directory.py => bzrlib/plugins/launchpad/test_lp_indirect.py
- bzrlib/tests/http_utils.py => bzrlib/tests/HTTPTestUtil.py
- bzrlib/tests/http_server.py => bzrlib/tests/HttpServer.py
- bzrlib/tests/test_revisionspec.py => bzrlib/tests/test_revisionnamespaces.py
- bzrlib/transport/ftp/__init__.py => bzrlib/transport/ftp.py
- bzrlib/xml8.py => bzrlib/xml5.py
- bzrlib/help_topics/en/authentication.txt => doc/en/user-guide/authentication_conf.txt
- doc/en/tutorials/centralized_workflow.txt => doc/en/user-guide/centralized_workflow.txt
- bzrlib/help_topics/en/configuration.txt => doc/en/user-guide/configuration.txt
- bzrlib/help_topics/en/conflicts.txt => doc/en/user-guide/conflicts.txt
- doc/en/tutorials/tutorial.txt => doc/en/user-guide/tutorial.txt
- bzrlib/help_topics/en/hooks.txt => doc/en/user-reference/hooks.txt
- files modified:
- .bzrignore
added
removed
bzrlib/knit.py
60
60
# record content length ?
61
61
62
62
63
from copy import copy
63
64
from cStringIO import StringIO
64
65
from itertools import izip, chain
65
66
import operator
66
67
import os
68
import sys
69
import warnings
70
from zlib import Z_DEFAULT_COMPRESSION
67
71
72
import bzrlib
68
73
from bzrlib.lazy_import import lazy_import
69
74
lazy_import(globals(), """
70
75
from bzrlib import (
71
76
annotate,
72
debug,
73
diff,
74
graph as _mod_graph,
75
index as _mod_index,
76
lru_cache,
77
77
pack,
78
progress,
79
78
trace,
80
tsort,
81
tuned_gzip,
82
79
)
83
80
""")
84
81
from bzrlib import (
82
cache_utf8,
83
debug,
84
diff,
85
85
errors,
86
86
osutils,
87
87
patiencediff,
88
progress,
89
merge,
90
ui,
88
91
)
89
92
from bzrlib.errors import (
90
93
FileExists,
92
95
KnitError,
93
96
InvalidRevisionId,
94
97
KnitCorrupt,
98
KnitDataStreamIncompatible,
95
99
KnitHeaderError,
96
100
RevisionNotPresent,
97
101
RevisionAlreadyPresent,
98
102
)
103
from bzrlib.tuned_gzip import GzipFile, bytes_to_gzip
99
104
from bzrlib.osutils import (
100
105
contains_whitespace,
101
106
contains_linebreaks,
102
107
sha_string,
103
108
sha_strings,
104
split_lines,
105
)
106
from bzrlib.versionedfile import (
107
AbsentContentFactory,
108
adapter_registry,
109
ConstantMapper,
110
ContentFactory,
111
FulltextContentFactory,
112
VersionedFile,
113
VersionedFiles,
114
)
109
)
110
from bzrlib.symbol_versioning import DEPRECATED_PARAMETER, deprecated_passed
111
from bzrlib.tsort import topo_sort
112
import bzrlib.ui
113
import bzrlib.weave
114
from bzrlib.versionedfile import VersionedFile, InterVersionedFile
115
115
116
116
117
117
# TODO: Split out code specific to this format into an associated object.
131
131
INDEX_SUFFIX = '.kndx'
132
132
133
133
134
class KnitAdapter(object):
135
"""Base class for knit record adaption."""
136
137
def __init__(self, basis_vf):
138
"""Create an adapter which accesses full texts from basis_vf.
139
140
:param basis_vf: A versioned file to access basis texts of deltas from.
141
May be None for adapters that do not need to access basis texts.
142
"""
143
self._data = KnitVersionedFiles(None, None)
144
self._annotate_factory = KnitAnnotateFactory()
145
self._plain_factory = KnitPlainFactory()
146
self._basis_vf = basis_vf
147
148
149
class FTAnnotatedToUnannotated(KnitAdapter):
150
"""An adapter from FT annotated knits to unannotated ones."""
151
152
def get_bytes(self, factory, annotated_compressed_bytes):
153
rec, contents = \
154
self._data._parse_record_unchecked(annotated_compressed_bytes)
155
content = self._annotate_factory.parse_fulltext(contents, rec[1])
156
size, bytes = self._data._record_to_data((rec[1],), rec[3], content.text())
157
return bytes
158
159
160
class DeltaAnnotatedToUnannotated(KnitAdapter):
161
"""An adapter for deltas from annotated to unannotated."""
162
163
def get_bytes(self, factory, annotated_compressed_bytes):
164
rec, contents = \
165
self._data._parse_record_unchecked(annotated_compressed_bytes)
166
delta = self._annotate_factory.parse_line_delta(contents, rec[1],
167
plain=True)
168
contents = self._plain_factory.lower_line_delta(delta)
169
size, bytes = self._data._record_to_data((rec[1],), rec[3], contents)
170
return bytes
171
172
173
class FTAnnotatedToFullText(KnitAdapter):
174
"""An adapter from FT annotated knits to unannotated ones."""
175
176
def get_bytes(self, factory, annotated_compressed_bytes):
177
rec, contents = \
178
self._data._parse_record_unchecked(annotated_compressed_bytes)
179
content, delta = self._annotate_factory.parse_record(factory.key[-1],
180
contents, factory._build_details, None)
181
return ''.join(content.text())
182
183
184
class DeltaAnnotatedToFullText(KnitAdapter):
185
"""An adapter for deltas from annotated to unannotated."""
186
187
def get_bytes(self, factory, annotated_compressed_bytes):
188
rec, contents = \
189
self._data._parse_record_unchecked(annotated_compressed_bytes)
190
delta = self._annotate_factory.parse_line_delta(contents, rec[1],
191
plain=True)
192
compression_parent = factory.parents[0]
193
basis_entry = self._basis_vf.get_record_stream(
194
[compression_parent], 'unordered', True).next()
195
if basis_entry.storage_kind == 'absent':
196
raise errors.RevisionNotPresent(compression_parent, self._basis_vf)
197
basis_lines = split_lines(basis_entry.get_bytes_as('fulltext'))
198
# Manually apply the delta because we have one annotated content and
199
# one plain.
200
basis_content = PlainKnitContent(basis_lines, compression_parent)
201
basis_content.apply_delta(delta, rec[1])
202
basis_content._should_strip_eol = factory._build_details[1]
203
return ''.join(basis_content.text())
204
205
206
class FTPlainToFullText(KnitAdapter):
207
"""An adapter from FT plain knits to unannotated ones."""
208
209
def get_bytes(self, factory, compressed_bytes):
210
rec, contents = \
211
self._data._parse_record_unchecked(compressed_bytes)
212
content, delta = self._plain_factory.parse_record(factory.key[-1],
213
contents, factory._build_details, None)
214
return ''.join(content.text())
215
216
217
class DeltaPlainToFullText(KnitAdapter):
218
"""An adapter for deltas from annotated to unannotated."""
219
220
def get_bytes(self, factory, compressed_bytes):
221
rec, contents = \
222
self._data._parse_record_unchecked(compressed_bytes)
223
delta = self._plain_factory.parse_line_delta(contents, rec[1])
224
compression_parent = factory.parents[0]
225
# XXX: string splitting overhead.
226
basis_entry = self._basis_vf.get_record_stream(
227
[compression_parent], 'unordered', True).next()
228
if basis_entry.storage_kind == 'absent':
229
raise errors.RevisionNotPresent(compression_parent, self._basis_vf)
230
basis_lines = split_lines(basis_entry.get_bytes_as('fulltext'))
231
basis_content = PlainKnitContent(basis_lines, compression_parent)
232
# Manually apply the delta because we have one annotated content and
233
# one plain.
234
content, _ = self._plain_factory.parse_record(rec[1], contents,
235
factory._build_details, basis_content)
236
return ''.join(content.text())
237
238
239
class KnitContentFactory(ContentFactory):
240
"""Content factory for streaming from knits.
241
242
:seealso ContentFactory:
243
"""
244
245
def __init__(self, key, parents, build_details, sha1, raw_record,
246
annotated, knit=None):
247
"""Create a KnitContentFactory for key.
248
249
:param key: The key.
250
:param parents: The parents.
251
:param build_details: The build details as returned from
252
get_build_details.
253
:param sha1: The sha1 expected from the full text of this object.
254
:param raw_record: The bytes of the knit data from disk.
255
:param annotated: True if the raw data is annotated.
256
"""
257
ContentFactory.__init__(self)
258
self.sha1 = sha1
259
self.key = key
260
self.parents = parents
261
if build_details[0] == 'line-delta':
262
kind = 'delta'
263
else:
264
kind = 'ft'
265
if annotated:
266
annotated_kind = 'annotated-'
267
else:
268
annotated_kind = ''
269
self.storage_kind = 'knit-%s%s-gz' % (annotated_kind, kind)
270
self._raw_record = raw_record
271
self._build_details = build_details
272
self._knit = knit
273
274
def get_bytes_as(self, storage_kind):
275
if storage_kind == self.storage_kind:
276
return self._raw_record
277
if storage_kind == 'fulltext' and self._knit is not None:
278
return self._knit.get_text(self.key[0])
279
else:
280
raise errors.UnavailableRepresentation(self.key, storage_kind,
281
self.storage_kind)
282
283
284
134
class KnitContent(object):
285
"""Content of a knit version to which deltas can be applied.
286
287
This is always stored in memory as a list of lines with \n at the end,
288
plus a flag saying if the final ending is really there or not, because that
289
corresponds to the on-disk knit representation.
290
"""
291
292
def __init__(self):
293
self._should_strip_eol = False
294
295
def apply_delta(self, delta, new_version_id):
296
"""Apply delta to this object to become new_version_id."""
297
raise NotImplementedError(self.apply_delta)
135
"""Content of a knit version to which deltas can be applied."""
136
137
def annotate(self):
138
"""Return a list of (origin, text) tuples."""
139
return list(self.annotate_iter())
298
140
299
141
def line_delta_iter(self, new_lines):
300
142
"""Generate line-based delta from this content to new_lines."""
341
183
"""Annotated content."""
342
184
343
185
def __init__(self, lines):
344
KnitContent.__init__(self)
345
186
self._lines = lines
346
187
347
def annotate(self):
348
"""Return a list of (origin, text) for each content line."""
349
lines = self._lines[:]
350
if self._should_strip_eol:
351
origin, last_line = lines[-1]
352
lines[-1] = (origin, last_line.rstrip('\n'))
353
return lines
188
def annotate_iter(self):
189
"""Yield tuples of (origin, text) for each content line."""
190
return iter(self._lines)
354
191
355
def apply_delta(self, delta, new_version_id):
356
"""Apply delta to this object to become new_version_id."""
357
offset = 0
358
lines = self._lines
359
for start, end, count, delta_lines in delta:
360
lines[offset+start:offset+end] = delta_lines
361
offset = offset + (start - end) + count
192
def strip_last_line_newline(self):
193
line = self._lines[-1][1].rstrip('\n')
194
self._lines[-1] = (self._lines[-1][0], line)
362
195
363
196
def text(self):
364
197
try:
365
lines = [text for origin, text in self._lines]
198
return [text for origin, text in self._lines]
366
199
except ValueError, e:
367
200
# most commonly (only?) caused by the internal form of the knit
368
201
# missing annotation information because of a bug - see thread
370
203
raise KnitCorrupt(self,
371
204
"line in annotated knit missing annotation information: %s"
372
205
% (e,))
373
if self._should_strip_eol:
374
lines[-1] = lines[-1].rstrip('\n')
375
return lines
376
206
377
207
def copy(self):
378
208
return AnnotatedKnitContent(self._lines[:])
387
217
"""
388
218
389
219
def __init__(self, lines, version_id):
390
KnitContent.__init__(self)
391
220
self._lines = lines
392
221
self._version_id = version_id
393
222
394
def annotate(self):
395
"""Return a list of (origin, text) for each content line."""
396
return [(self._version_id, line) for line in self._lines]
397
398
def apply_delta(self, delta, new_version_id):
399
"""Apply delta to this object to become new_version_id."""
400
offset = 0
401
lines = self._lines
402
for start, end, count, delta_lines in delta:
403
lines[offset+start:offset+end] = delta_lines
404
offset = offset + (start - end) + count
405
self._version_id = new_version_id
223
def annotate_iter(self):
224
"""Yield tuples of (origin, text) for each content line."""
225
for line in self._lines:
226
yield self._version_id, line
406
227
407
228
def copy(self):
408
229
return PlainKnitContent(self._lines[:], self._version_id)
409
230
231
def strip_last_line_newline(self):
232
self._lines[-1] = self._lines[-1].rstrip('\n')
233
410
234
def text(self):
411
lines = self._lines
412
if self._should_strip_eol:
413
lines = lines[:]
414
lines[-1] = lines[-1].rstrip('\n')
415
return lines
416
417
418
class _KnitFactory(object):
419
"""Base class for common Factory functions."""
420
421
def parse_record(self, version_id, record, record_details,
422
base_content, copy_base_content=True):
423
"""Parse a record into a full content object.
424
425
:param version_id: The official version id for this content
426
:param record: The data returned by read_records_iter()
427
:param record_details: Details about the record returned by
428
get_build_details
429
:param base_content: If get_build_details returns a compression_parent,
430
you must return a base_content here, else use None
431
:param copy_base_content: When building from the base_content, decide
432
you can either copy it and return a new object, or modify it in
433
place.
434
:return: (content, delta) A Content object and possibly a line-delta,
435
delta may be None
436
"""
437
method, noeol = record_details
438
if method == 'line-delta':
439
if copy_base_content:
440
content = base_content.copy()
441
else:
442
content = base_content
443
delta = self.parse_line_delta(record, version_id)
444
content.apply_delta(delta, version_id)
445
else:
446
content = self.parse_fulltext(record, version_id)
447
delta = None
448
content._should_strip_eol = noeol
449
return (content, delta)
450
451
452
class KnitAnnotateFactory(_KnitFactory):
235
return self._lines
236
237
238
class KnitAnnotateFactory(object):
453
239
"""Factory for creating annotated Content objects."""
454
240
455
241
annotated = True
556
342
for origin, text in lines)
557
343
return out
558
344
559
def annotate(self, knit, key):
560
content = knit._get_content(key)
561
# adjust for the fact that serialised annotations are only key suffixes
562
# for this factory.
563
if type(key) == tuple:
564
prefix = key[:-1]
565
origins = content.annotate()
566
result = []
567
for origin, line in origins:
568
result.append((prefix + (origin,), line))
569
return result
570
else:
571
# XXX: This smells a bit. Why would key ever be a non-tuple here?
572
# Aren't keys defined to be tuples? -- spiv 20080618
573
return content.annotate()
574
575
576
class KnitPlainFactory(_KnitFactory):
345
def annotate_iter(self, knit, version_id):
346
content = knit._get_content(version_id)
347
return content.annotate_iter()
348
349
350
class KnitPlainFactory(object):
577
351
"""Factory for creating plain Content objects."""
578
352
579
353
annotated = False
630
404
out.extend(lines)
631
405
return out
632
406
633
def annotate(self, knit, key):
634
annotator = _KnitAnnotator(knit)
635
return annotator.annotate(key)
636
637
638
639
def make_file_factory(annotated, mapper):
640
"""Create a factory for creating a file based KnitVersionedFiles.
641
642
This is only functional enough to run interface tests, it doesn't try to
643
provide a full pack environment.
644
645
:param annotated: knit annotations are wanted.
646
:param mapper: The mapper from keys to paths.
647
"""
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)
652
return factory
653
654
655
def make_pack_factory(graph, delta, keylength):
656
"""Create a factory for creating a pack based VersionedFiles.
657
658
This is only functional enough to run interface tests, it doesn't try to
659
provide a full pack environment.
660
661
:param graph: Store a graph.
662
:param delta: Delta compress contents.
663
:param keylength: How long should keys be.
664
"""
665
def factory(transport):
666
parents = graph or delta
667
ref_length = 0
668
if graph:
669
ref_length += 1
407
def annotate_iter(self, knit, version_id):
408
return annotate_knit(knit, version_id)
409
410
411
def make_empty_knit(transport, relpath):
412
"""Construct a empty knit at the specified location."""
413
k = KnitVersionedFile(transport, relpath, 'w', KnitPlainFactory)
414
415
416
class KnitVersionedFile(VersionedFile):
417
"""Weave-like structure with faster random access.
418
419
A knit stores a number of texts and a summary of the relationships
420
between them. Texts are identified by a string version-id. Texts
421
are normally stored and retrieved as a series of lines, but can
422
also be passed as single strings.
423
424
Lines are stored with the trailing newline (if any) included, to
425
avoid special cases for files with no final newline. Lines are
426
composed of 8-bit characters, not unicode. The combination of
427
these approaches should mean any 'binary' file can be safely
428
stored and retrieved.
429
"""
430
431
def __init__(self, relpath, transport, file_mode=None, access_mode=None,
432
factory=None, delta=True, create=False, create_parent_dir=False,
433
delay_create=False, dir_mode=None, index=None, access_method=None):
434
"""Construct a knit at location specified by relpath.
435
436
:param create: If not True, only open an existing knit.
437
:param create_parent_dir: If True, create the parent directory if
438
creating the file fails. (This is used for stores with
439
hash-prefixes that may not exist yet)
440
:param delay_create: The calling code is aware that the knit won't
441
actually be created until the first data is stored.
442
:param index: An index to use for the knit.
443
"""
444
if access_mode is None:
445
access_mode = 'w'
446
super(KnitVersionedFile, self).__init__(access_mode)
447
assert access_mode in ('r', 'w'), "invalid mode specified %r" % access_mode
448
self.transport = transport
449
self.filename = relpath
450
self.factory = factory or KnitAnnotateFactory()
451
self.writable = (access_mode == 'w')
452
self.delta = delta
453
454
self._max_delta_chain = 200
455
456
if index is None:
457
self._index = _KnitIndex(transport, relpath + INDEX_SUFFIX,
458
access_mode, create=create, file_mode=file_mode,
459
create_parent_dir=create_parent_dir, delay_create=delay_create,
460
dir_mode=dir_mode)
461
else:
462
self._index = index
463
if access_method is None:
464
_access = _KnitAccess(transport, relpath + DATA_SUFFIX, file_mode, dir_mode,
465
((create and not len(self)) and delay_create), create_parent_dir)
466
else:
467
_access = access_method
468
if create and not len(self) and not delay_create:
469
_access.create()
470
self._data = _KnitData(_access)
471
472
def __repr__(self):
473
return '%s(%s)' % (self.__class__.__name__,
474
self.transport.abspath(self.filename))
475
476
def _check_should_delta(self, first_parents):
477
"""Iterate back through the parent listing, looking for a fulltext.
478
479
This is used when we want to decide whether to add a delta or a new
480
fulltext. It searches for _max_delta_chain parents. When it finds a
481
fulltext parent, it sees if the total size of the deltas leading up to
482
it is large enough to indicate that we want a new full text anyway.
483
484
Return True if we should create a new delta, False if we should use a
485
full text.
486
"""
487
delta_size = 0
488
fulltext_size = None
489
delta_parents = first_parents
490
for count in xrange(self._max_delta_chain):
491
parent = delta_parents[0]
492
method = self._index.get_method(parent)
493
index, pos, size = self._index.get_position(parent)
494
if method == 'fulltext':
495
fulltext_size = size
496
break
497
delta_size += size
498
delta_parents = self._index.get_parents(parent)
499
else:
500
# We couldn't find a fulltext, so we must create a new one
501
return False
502
503
return fulltext_size > delta_size
504
505
def _add_raw_records(self, records, data):
506
"""Add all the records 'records' with data pre-joined in 'data'.
507
508
:param records: A list of tuples(version_id, options, parents, size).
509
:param data: The data for the records. When it is written, the records
510
are adjusted to have pos pointing into data by the sum of
511
the preceding records sizes.
512
"""
513
# write all the data
514
raw_record_sizes = [record[3] for record in records]
515
positions = self._data.add_raw_records(raw_record_sizes, data)
516
offset = 0
517
index_entries = []
518
for (version_id, options, parents, size), access_memo in zip(
519
records, positions):
520
index_entries.append((version_id, options, access_memo, parents))
521
if self._data._do_cache:
522
self._data._cache[version_id] = data[offset:offset+size]
523
offset += size
524
self._index.add_versions(index_entries)
525
526
def enable_cache(self):
527
"""Start caching data for this knit"""
528
self._data.enable_cache()
529
530
def clear_cache(self):
531
"""Clear the data cache only."""
532
self._data.clear_cache()
533
534
def copy_to(self, name, transport):
535
"""See VersionedFile.copy_to()."""
536
# copy the current index to a temp index to avoid racing with local
537
# writes
538
transport.put_file_non_atomic(name + INDEX_SUFFIX + '.tmp',
539
self.transport.get(self._index._filename))
540
# copy the data file
541
f = self._data._open_file()
542
try:
543
transport.put_file(name + DATA_SUFFIX, f)
544
finally:
545
f.close()
546
# move the copied index into place
547
transport.move(name + INDEX_SUFFIX + '.tmp', name + INDEX_SUFFIX)
548
549
def create_empty(self, name, transport, mode=None):
550
return KnitVersionedFile(name, transport, factory=self.factory,
551
delta=self.delta, create=True)
552
553
def get_data_stream(self, required_versions):
554
"""Get a data stream for the specified versions.
555
556
Versions may be returned in any order, not necessarily the order
557
specified.
558
559
:param required_versions: The exact set of versions to be extracted.
560
Unlike some other knit methods, this is not used to generate a
561
transitive closure, rather it is used precisely as given.
562
563
:returns: format_signature, list of (version, options, length, parents),
564
reader_callable.
565
"""
566
if not isinstance(required_versions, set):
567
required_versions = set(required_versions)
568
# we don't care about inclusions, the caller cares.
569
# but we need to setup a list of records to visit.
570
for version_id in required_versions:
571
if not self.has_version(version_id):
572
raise RevisionNotPresent(version_id, self.filename)
573
# Pick the desired versions out of the index in oldest-to-newest order
574
version_list = []
575
for version_id in self.versions():
576
if version_id in required_versions:
577
version_list.append(version_id)
578
579
# create the list of version information for the result
580
copy_queue_records = []
581
copy_set = set()
582
result_version_list = []
583
for version_id in version_list:
584
options = self._index.get_options(version_id)
585
parents = self._index.get_parents_with_ghosts(version_id)
586
index_memo = self._index.get_position(version_id)
587
copy_queue_records.append((version_id, index_memo))
588
none, data_pos, data_size = index_memo
589
copy_set.add(version_id)
590
# version, options, length, parents
591
result_version_list.append((version_id, options, data_size,
592
parents))
593
594
# Read the compressed record data.
595
# XXX:
596
# From here down to the return should really be logic in the returned
597
# callable -- in a class that adapts read_records_iter_raw to read
598
# requests.
599
raw_datum = []
600
for (version_id, raw_data), \
601
(version_id2, options, _, parents) in \
602
izip(self._data.read_records_iter_raw(copy_queue_records),
603
result_version_list):
604
assert version_id == version_id2, 'logic error, inconsistent results'
605
raw_datum.append(raw_data)
606
pseudo_file = StringIO(''.join(raw_datum))
607
def read(length):
608
if length is None:
609
return pseudo_file.read()
610
else:
611
return pseudo_file.read(length)
612
return (self.get_format_signature(), result_version_list, read)
613
614
def _extract_blocks(self, version_id, source, target):
615
if self._index.get_method(version_id) != 'line-delta':
616
return None
617
parent, sha1, noeol, delta = self.get_delta(version_id)
618
return KnitContent.get_line_delta_blocks(delta, source, target)
619
620
def get_delta(self, version_id):
621
"""Get a delta for constructing version from some other version."""
622
self.check_not_reserved_id(version_id)
623
parents = self.get_parents(version_id)
624
if len(parents):
625
parent = parents[0]
626
else:
627
parent = None
628
index_memo = self._index.get_position(version_id)
629
data, sha1 = self._data.read_records(((version_id, index_memo),))[version_id]
630
noeol = 'no-eol' in self._index.get_options(version_id)
631
if 'fulltext' == self._index.get_method(version_id):
632
new_content = self.factory.parse_fulltext(data, version_id)
633
if parent is not None:
634
reference_content = self._get_content(parent)
635
old_texts = reference_content.text()
636
else:
637
old_texts = []
638
new_texts = new_content.text()
639
delta_seq = patiencediff.PatienceSequenceMatcher(None, old_texts,
640
new_texts)
641
return parent, sha1, noeol, self._make_line_delta(delta_seq, new_content)
642
else:
643
delta = self.factory.parse_line_delta(data, version_id)
644
return parent, sha1, noeol, delta
645
646
def get_format_signature(self):
647
"""See VersionedFile.get_format_signature()."""
648
if self.factory.annotated:
649
annotated_part = "annotated"
650
else:
651
annotated_part = "plain"
652
return "knit-%s" % (annotated_part,)
653
654
def get_graph_with_ghosts(self):
655
"""See VersionedFile.get_graph_with_ghosts()."""
656
graph_items = self._index.get_graph()
657
return dict(graph_items)
658
659
def get_sha1(self, version_id):
660
return self.get_sha1s([version_id])[0]
661
662
def get_sha1s(self, version_ids):
663
"""See VersionedFile.get_sha1()."""
664
record_map = self._get_record_map(version_ids)
665
# record entry 2 is the 'digest'.
666
return [record_map[v][2] for v in version_ids]
667
668
@staticmethod
669
def get_suffixes():
670
"""See VersionedFile.get_suffixes()."""
671
return [DATA_SUFFIX, INDEX_SUFFIX]
672
673
def has_ghost(self, version_id):
674
"""True if there is a ghost reference in the file to version_id."""
675
# maybe we have it
676
if self.has_version(version_id):
677
return False
678
# optimisable if needed by memoising the _ghosts set.
679
items = self._index.get_graph()
680
for node, parents in items:
681
for parent in parents:
682
if parent not in self._index._cache:
683
if parent == version_id:
684
return True
685
return False
686
687
def insert_data_stream(self, (format, data_list, reader_callable)):
688
"""Insert knit records from a data stream into this knit.
689
690
If a version in the stream is already present in this knit, it will not
691
be inserted a second time. It will be checked for consistency with the
692
stored version however, and may cause a KnitCorrupt error to be raised
693
if the data in the stream disagrees with the already stored data.
694
695
:seealso: get_data_stream
696
"""
697
if format != self.get_format_signature():
698
trace.mutter('incompatible format signature inserting to %r', self)
699
raise KnitDataStreamIncompatible(
700
format, self.get_format_signature())
701
702
for version_id, options, length, parents in data_list:
703
if self.has_version(version_id):
704
# First check: the list of parents.
705
my_parents = self.get_parents_with_ghosts(version_id)
706
if my_parents != parents:
707
# XXX: KnitCorrupt is not quite the right exception here.
708
raise KnitCorrupt(
709
self.filename,
710
'parents list %r from data stream does not match '
711
'already recorded parents %r for %s'
712
% (parents, my_parents, version_id))
713
714
# Also check the SHA-1 of the fulltext this content will
715
# produce.
716
raw_data = reader_callable(length)
717
my_fulltext_sha1 = self.get_sha1(version_id)
718
df, rec = self._data._parse_record_header(version_id, raw_data)
719
stream_fulltext_sha1 = rec[3]
720
if my_fulltext_sha1 != stream_fulltext_sha1:
721
# Actually, we don't know if it's this knit that's corrupt,
722
# or the data stream we're trying to insert.
723
raise KnitCorrupt(
724
self.filename, 'sha-1 does not match %s' % version_id)
725
else:
726
if 'line-delta' in options:
727
# Make sure that this knit record is actually useful: a
728
# line-delta is no use unless we have its parent.
729
# Fetching from a broken repository with this problem
730
# shouldn't break the target repository.
731
if not self._index.has_version(parents[0]):
732
raise KnitCorrupt(
733
self.filename,
734
'line-delta from stream references '
735
'missing parent %s' % parents[0])
736
self._add_raw_records(
737
[(version_id, options, parents, length)],
738
reader_callable(length))
739
740
def versions(self):
741
"""See VersionedFile.versions."""
742
if 'evil' in debug.debug_flags:
743
trace.mutter_callsite(2, "versions scales with size of history")
744
return self._index.get_versions()
745
746
def has_version(self, version_id):
747
"""See VersionedFile.has_version."""
748
if 'evil' in debug.debug_flags:
749
trace.mutter_callsite(2, "has_version is a LBYL scenario")
750
return self._index.has_version(version_id)
751
752
__contains__ = has_version
753
754
def _merge_annotations(self, content, parents, parent_texts={},
755
delta=None, annotated=None,
756
left_matching_blocks=None):
757
"""Merge annotations for content. This is done by comparing
758
the annotations based on changed to the text.
759
"""
760
if left_matching_blocks is not None:
761
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
762
else:
763
delta_seq = None
764
if annotated:
765
for parent_id in parents:
766
merge_content = self._get_content(parent_id, parent_texts)
767
if (parent_id == parents[0] and delta_seq is not None):
768
seq = delta_seq
769
else:
770
seq = patiencediff.PatienceSequenceMatcher(
771
None, merge_content.text(), content.text())
772
for i, j, n in seq.get_matching_blocks():
773
if n == 0:
774
continue
775
# this appears to copy (origin, text) pairs across to the
776
# new content for any line that matches the last-checked
777
# parent.
778
content._lines[j:j+n] = merge_content._lines[i:i+n]
670
779
if delta:
671
ref_length += 1
672
max_delta_chain = 200
673
else:
674
max_delta_chain = 0
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)
679
writer.begin()
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
688
return result
689
return factory
690
691
692
def cleanup_pack_knit(versioned_files):
693
versioned_files.stream.close()
694
versioned_files.writer.end()
695
696
697
class KnitVersionedFiles(VersionedFiles):
698
"""Storage for many versioned files using knit compression.
699
700
Backend storage is managed by indices and data objects.
701
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.
706
"""
707
708
def __init__(self, index, data_access, max_delta_chain=200,
709
annotated=False):
710
"""Create a KnitVersionedFiles with index and data_access.
711
712
:param index: The index for the knit data.
713
:param data_access: The access object to store and retrieve knit
714
records.
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.
719
"""
720
self._index = index
721
self._access = data_access
722
self._max_delta_chain = max_delta_chain
723
if annotated:
724
self._factory = KnitAnnotateFactory()
725
else:
726
self._factory = KnitPlainFactory()
727
self._fallback_vfs = []
728
729
def add_fallback_versioned_files(self, a_versioned_files):
730
"""Add a source of texts for texts not present in this knit.
731
732
:param a_versioned_files: A VersionedFiles object.
733
"""
734
self._fallback_vfs.append(a_versioned_files)
735
736
def add_lines(self, key, parents, lines, parent_texts=None,
737
left_matching_blocks=None, nostore_sha=None, random_id=False,
738
check_content=True):
739
"""See VersionedFiles.add_lines()."""
740
self._index._check_write_ok()
741
self._check_add(key, lines, random_id, check_content)
742
if parents is None:
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.
746
parents = ()
747
return self._add(key, lines, parents,
780
if delta_seq is None:
781
reference_content = self._get_content(parents[0], parent_texts)
782
new_texts = content.text()
783
old_texts = reference_content.text()
784
delta_seq = patiencediff.PatienceSequenceMatcher(
785
None, old_texts, new_texts)
786
return self._make_line_delta(delta_seq, content)
787
788
def _make_line_delta(self, delta_seq, new_content):
789
"""Generate a line delta from delta_seq and new_content."""
790
diff_hunks = []
791
for op in delta_seq.get_opcodes():
792
if op[0] == 'equal':
793
continue
794
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
795
return diff_hunks
796
797
def _get_components_positions(self, version_ids):
798
"""Produce a map of position data for the components of versions.
799
800
This data is intended to be used for retrieving the knit records.
801
802
A dict of version_id to (method, data_pos, data_size, next) is
803
returned.
804
method is the way referenced data should be applied.
805
data_pos is the position of the data in the knit.
806
data_size is the size of the data in the knit.
807
next is the build-parent of the version, or None for fulltexts.
808
"""
809
component_data = {}
810
for version_id in version_ids:
811
cursor = version_id
812
813
while cursor is not None and cursor not in component_data:
814
method = self._index.get_method(cursor)
815
if method == 'fulltext':
816
next = None
817
else:
818
next = self.get_parents(cursor)[0]
819
index_memo = self._index.get_position(cursor)
820
component_data[cursor] = (method, index_memo, next)
821
cursor = next
822
return component_data
823
824
def _get_content(self, version_id, parent_texts={}):
825
"""Returns a content object that makes up the specified
826
version."""
827
cached_version = parent_texts.get(version_id, None)
828
if cached_version is not None:
829
if not self.has_version(version_id):
830
raise RevisionNotPresent(version_id, self.filename)
831
return cached_version
832
833
text_map, contents_map = self._get_content_maps([version_id])
834
return contents_map[version_id]
835
836
def _check_versions_present(self, version_ids):
837
"""Check that all specified versions are present."""
838
self._index.check_versions_present(version_ids)
839
840
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
841
nostore_sha, random_id, check_content):
842
"""See VersionedFile.add_lines_with_ghosts()."""
843
self._check_add(version_id, lines, random_id, check_content)
844
return self._add(version_id, lines, parents, self.delta,
845
parent_texts, None, nostore_sha, random_id)
846
847
def _add_lines(self, version_id, parents, lines, parent_texts,
848
left_matching_blocks, nostore_sha, random_id, check_content):
849
"""See VersionedFile.add_lines."""
850
self._check_add(version_id, lines, random_id, check_content)
851
self._check_versions_present(parents)
852
return self._add(version_id, lines[:], parents, self.delta,
748
853
parent_texts, left_matching_blocks, nostore_sha, random_id)
749
854
750
def _add(self, key, lines, parents, parent_texts,
855
def _check_add(self, version_id, lines, random_id, check_content):
856
"""check that version_id and lines are safe to add."""
857
if contains_whitespace(version_id):
858
raise InvalidRevisionId(version_id, self.filename)
859
self.check_not_reserved_id(version_id)
860
# Technically this could be avoided if we are happy to allow duplicate
861
# id insertion when other things than bzr core insert texts, but it
862
# seems useful for folk using the knit api directly to have some safety
863
# blanket that we can disable.
864
if not random_id and self.has_version(version_id):
865
raise RevisionAlreadyPresent(version_id, self.filename)
866
if check_content:
867
self._check_lines_not_unicode(lines)
868
self._check_lines_are_lines(lines)
869
870
def _add(self, version_id, lines, parents, delta, parent_texts,
751
871
left_matching_blocks, nostore_sha, random_id):
752
872
"""Add a set of lines on top of version specified by parents.
753
873
874
If delta is true, compress the text as a line-delta against
875
the first parent.
876
754
877
Any versions not present will be converted into ghosts.
755
878
"""
756
879
# first thing, if the content is something we don't need to store, find
763
886
present_parents = []
764
887
if parent_texts is None:
765
888
parent_texts = {}
766
# Do a single query to ascertain parent presence.
767
present_parent_map = self.get_parent_map(parents)
768
889
for parent in parents:
769
if parent in present_parent_map:
890
if self.has_version(parent):
770
891
present_parents.append(parent)
771
892
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]):
893
# can only compress against the left most present parent.
894
if (delta and
895
(len(present_parents) == 0 or
896
present_parents[0] != parents[0])):
775
897
delta = False
776
else:
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])
781
898
782
899
text_length = len(line_bytes)
783
900
options = []
789
906
lines[-1] = lines[-1] + '\n'
790
907
line_bytes += '\n'
791
908
792
for element in key:
793
if type(element) != str:
794
raise TypeError("key contains non-strings: %r" % (key,))
795
# Knit hunks are still last-element only
796
version_id = key[-1]
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
800
# EOL.
801
content._should_strip_eol = True
802
if delta or (self._factory.annotated and len(present_parents) > 0):
909
if delta:
910
# To speed the extract of texts the delta chain is limited
911
# to a fixed number of deltas. This should minimize both
912
# I/O and the time spend applying deltas.
913
delta = self._check_should_delta(present_parents)
914
915
assert isinstance(version_id, str)
916
content = self.factory.make(lines, version_id)
917
if delta or (self.factory.annotated and len(present_parents) > 0):
803
918
# Merge annotations from parent texts if needed.
804
919
delta_hunks = self._merge_annotations(content, present_parents,
805
parent_texts, delta, self._factory.annotated,
920
parent_texts, delta, self.factory.annotated,
806
921
left_matching_blocks)
807
922
808
923
if delta:
809
924
options.append('line-delta')
810
store_lines = self._factory.lower_line_delta(delta_hunks)
811
size, bytes = self._record_to_data(key, digest,
925
store_lines = self.factory.lower_line_delta(delta_hunks)
926
size, bytes = self._data._record_to_data(version_id, digest,
812
927
store_lines)
813
928
else:
814
929
options.append('fulltext')
815
930
# isinstance is slower and we have no hierarchy.
816
if self._factory.__class__ == KnitPlainFactory:
931
if self.factory.__class__ == KnitPlainFactory:
817
932
# Use the already joined bytes saving iteration time in
818
933
# _record_to_data.
819
size, bytes = self._record_to_data(key, digest,
934
size, bytes = self._data._record_to_data(version_id, digest,
820
935
lines, [line_bytes])
821
936
else:
822
937
# get mixed annotation + content and feed it into the
823
938
# serialiser.
824
store_lines = self._factory.lower_fulltext(content)
825
size, bytes = self._record_to_data(key, digest,
939
store_lines = self.factory.lower_fulltext(content)
940
size, bytes = self._data._record_to_data(version_id, digest,
826
941
store_lines)
827
942
828
access_memo = self._access.add_raw_records([(key, size)], bytes)[0]
829
self._index.add_records(
830
((key, options, access_memo, parents),),
943
access_memo = self._data.add_raw_records([size], bytes)[0]
944
self._index.add_versions(
945
((version_id, options, access_memo, parents),),
831
946
random_id=random_id)
832
947
return digest, text_length, content
833
948
834
def annotate(self, key):
835
"""See VersionedFiles.annotate."""
836
return self._factory.annotate(self, key)
837
838
949
def check(self, progress_bar=None):
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
843
# no basis.
844
keys = self._index.keys()
845
parent_map = self.get_parent_map(keys)
846
for key in 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:
854
fallback_vfs.check()
855
856
def _check_add(self, key, lines, random_id, check_content):
857
"""check that version_id and lines are safe to add."""
858
version_id = key[-1]
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.
866
if check_content:
867
self._check_lines_not_unicode(lines)
868
self._check_lines_are_lines(lines)
869
870
def _check_header(self, key, line):
871
rec = self._split_header(line)
872
self._check_header_version(rec, key[-1])
873
return rec
874
875
def _check_header_version(self, rec, version_id):
876
"""Checks the header version on original format knit records.
950
"""See VersionedFile.check()."""
951
952
def _clone_text(self, new_version_id, old_version_id, parents):
953
"""See VersionedFile.clone_text()."""
954
# FIXME RBC 20060228 make fast by only inserting an index with null
955
# delta.
956
self.add_lines(new_version_id, parents, self.get_lines(old_version_id))
957
958
def get_lines(self, version_id):
959
"""See VersionedFile.get_lines()."""
960
return self.get_line_list([version_id])[0]
961
962
def _get_record_map(self, version_ids):
963
"""Produce a dictionary of knit records.
877
964
878
These have the last component of the key embedded in the record.
879
"""
880
if rec[1] != version_id:
881
raise KnitCorrupt(self,
882
'unexpected version, wanted %r, got %r' % (version_id, rec[1]))
883
884
def _check_should_delta(self, parent):
885
"""Iterate back through the parent listing, looking for a fulltext.
886
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.
891
892
Return True if we should create a new delta, False if we should use a
893
full text.
894
"""
895
delta_size = 0
896
fulltext_size = None
897
for count in xrange(self._max_delta_chain):
898
# XXX: Collapse these two queries:
899
try:
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
903
# boundaries.
904
method = self._index.get_method(parent)
905
except RevisionNotPresent:
906
# Some basis is not locally present: always delta
907
return False
908
index, pos, size = self._index.get_position(parent)
909
if method == 'fulltext':
910
fulltext_size = size
911
break
912
delta_size += size
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
916
# parent.
917
parent = self._index.get_parent_map([parent])[parent][0]
918
else:
919
# We couldn't find a fulltext, so we must create a new one
920
return False
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
924
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]
929
930
def _get_components_positions(self, keys, allow_missing=False):
931
"""Produce a map of position data for the components of keys.
932
933
This data is intended to be used for retrieving the knit records.
934
935
A dict of key to (record_details, index_memo, next, parents) is
936
returned.
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
939
data
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
942
943
:param allow_missing: If True do not raise an error on a missing component,
944
just ignore it.
945
"""
946
component_data = {}
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
963
964
def _get_content(self, key, parent_texts={}):
965
"""Returns a content object that makes up the specified
966
version."""
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]
975
976
def _get_content_maps(self, keys, nonlocal_keys=None):
965
The keys are version_ids, the values are tuples of (method, content,
966
digest, next).
967
method is the way the content should be applied.
968
content is a KnitContent object.
969
digest is the SHA1 digest of this version id after all steps are done
970
next is the build-parent of the version, i.e. the leftmost ancestor.
971
If the method is fulltext, next will be None.
972
"""
973
position_map = self._get_components_positions(version_ids)
974
# c = component_id, m = method, i_m = index_memo, n = next
975
records = [(c, i_m) for c, (m, i_m, n) in position_map.iteritems()]
976
record_map = {}
977
for component_id, content, digest in \
978
self._data.read_records_iter(records):
979
method, index_memo, next = position_map[component_id]
980
record_map[component_id] = method, content, digest, next
981
982
return record_map
983
984
def get_text(self, version_id):
985
"""See VersionedFile.get_text"""
986
return self.get_texts([version_id])[0]
987
988
def get_texts(self, version_ids):
989
return [''.join(l) for l in self.get_line_list(version_ids)]
990
991
def get_line_list(self, version_ids):
992
"""Return the texts of listed versions as a list of strings."""
993
for version_id in version_ids:
994
self.check_not_reserved_id(version_id)
995
text_map, content_map = self._get_content_maps(version_ids)
996
return [text_map[v] for v in version_ids]
997
998
_get_lf_split_line_list = get_line_list
999
1000
def _get_content_maps(self, version_ids):
977
1001
"""Produce maps of text and KnitContents
978
1002
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
982
fallback knits.
983
1003
:return: (text_map, content_map) where text_map contains the texts for
984
the requested versions and content_map contains the KnitContents.
1004
the requested versions and content_map contains the KnitContents.
1005
Both dicts take version_ids as their keys.
985
1006
"""
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
989
# final output.
990
keys = list(keys)
991
multiple_versions = len(keys) != 1
992
record_map = self._get_record_map(keys, allow_missing=True)
1007
record_map = self._get_record_map(version_ids)
993
1008
994
1009
text_map = {}
995
1010
content_map = {}
996
1011
final_content = {}
997
if nonlocal_keys is None:
998
nonlocal_keys = set()
999
else:
1000
nonlocal_keys = frozenset(nonlocal_keys)
1001
missing_keys = set(nonlocal_keys)
1002
for source in self._fallback_vfs:
1003
if not missing_keys:
1004
break
1005
for record in source.get_record_stream(missing_keys,
1006
'unordered', True):
1007
if record.storage_kind == 'absent':
1008
continue
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]
1015
for key in keys:
1016
if key in nonlocal_keys:
1017
# already handled
1018
continue
1012
for version_id in version_ids:
1019
1013
components = []
1020
cursor = key
1014
cursor = version_id
1021
1015
while cursor is not None:
1022
try:
1023
record, record_details, digest, next = record_map[cursor]
1024
except KeyError:
1025
raise RevisionNotPresent(cursor, self)
1026
components.append((cursor, record, record_details, digest))
1027
cursor = next
1016
method, data, digest, next = record_map[cursor]
1017
components.append((cursor, method, data, digest))
1028
1018
if cursor in content_map:
1029
# no need to plan further back
1030
components.append((cursor, None, None, None))
1031
1019
break
1020
cursor = next
1032
1021
1033
1022
content = None
1034
for (component_id, record, record_details,
1035
digest) in reversed(components):
1023
for component_id, method, data, digest in reversed(components):
1036
1024
if component_id in content_map:
1037
1025
content = content_map[component_id]
1038
1026
else:
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
1027
if method == 'fulltext':
1028
assert content is None
1029
content = self.factory.parse_fulltext(data, version_id)
1030
elif method == 'line-delta':
1031
delta = self.factory.parse_line_delta(data, version_id)
1032
content = content.copy()
1033
content._lines = self._apply_delta(content._lines,
1034
delta)
1035
content_map[component_id] = content
1044
1036
1045
final_content[key] = content
1037
if 'no-eol' in self._index.get_options(version_id):
1038
content = content.copy()
1039
content.strip_last_line_newline()
1040
final_content[version_id] = content
1046
1041
1047
1042
# digest here is the digest from the last applied component.
1048
1043
text = content.text()
1049
1044
actual_sha = sha_strings(text)
1050
1045
if actual_sha != digest:
1051
raise KnitCorrupt(self,
1046
raise KnitCorrupt(self.filename,
1052
1047
'\n sha-1 %s'
1053
1048
'\n of reconstructed text does not match'
1054
1049
'\n expected %s'
1055
1050
'\n for version %s' %
1056
(actual_sha, digest, key))
1057
text_map[key] = text
1051
(actual_sha, digest, version_id))
1052
text_map[version_id] = text
1058
1053
return text_map, final_content
1059
1054
1060
def get_parent_map(self, keys):
1061
"""Get a map of the graph parents of keys.
1062
1063
:param keys: The keys to look up parents for.
1064
:return: A mapping from keys to parents. Absent keys are absent from
1065
the mapping.
1066
"""
1067
return self._get_parent_map_with_sources(keys)[0]
1068
1069
def _get_parent_map_with_sources(self, keys):
1070
"""Get a map of the parents of keys.
1071
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,
1077
and so on.
1078
"""
1079
result = {}
1080
sources = [self._index] + self._fallback_vfs
1081
source_results = []
1082
missing = set(keys)
1083
for source in sources:
1084
if not missing:
1085
break
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
1091
1092
def _get_record_map(self, keys, allow_missing=False):
1093
"""Produce a dictionary of knit records.
1094
1095
:return: {key:(record, record_details, digest, next)}
1096
record
1097
data returned from read_records
1098
record_details
1099
opaque information to pass to parse_record
1100
digest
1101
SHA1 digest of the full text after all steps are done
1102
next
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.
1108
"""
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()]
1114
record_map = {}
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
1119
return record_map
1120
1121
def get_record_stream(self, keys, ordering, include_delta_closure):
1122
"""Get a stream of records for keys.
1123
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
1127
children.
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.
1132
"""
1133
# keys might be a generator
1134
keys = set(keys)
1135
if not keys:
1136
return
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)
1142
else:
1143
build_details = self._index.get_build_details(keys)
1144
# map from key to
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 = {}
1156
for key in keys:
1157
# the delta chain
1158
try:
1159
chain = [key, positions[key][2]]
1160
except KeyError:
1161
needed_from_fallback.add(key)
1162
continue
1163
result = True
1164
while chain[-1] is not None:
1165
if chain[-1] in reconstructable_keys:
1166
result = reconstructable_keys[chain[-1]]
1167
break
1168
else:
1169
try:
1170
chain.append(positions[chain[-1]][2])
1171
except KeyError:
1172
# missing basis component
1173
needed_from_fallback.add(chain[-1])
1174
result = True
1175
break
1176
for chain_key in chain[:-1]:
1177
reconstructable_keys[chain_key] = result
1178
if not 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:
1186
source_keys = []
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
1192
break
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)
1197
else:
1198
if ordering != 'unordered':
1199
raise AssertionError('valid values for ordering are:'
1200
' "unordered" or "topological" not: %r'
1201
% (ordering,))
1202
# Just group by source; remote sources first.
1203
present_keys = []
1204
source_keys = []
1205
for parent_map in reversed(parent_maps):
1206
source_keys.append((parent_map, []))
1207
for key in parent_map:
1208
present_keys.append(key)
1209
source_keys[-1][1].append(key)
1210
absent_keys = keys - set(global_map)
1211
for key in absent_keys:
1212
yield AbsentContentFactory(key)
1213
# restrict our view to the keys we can answer.
1214
# XXX: Memory: TODO: batch data here to cap buffered data at (say) 1MB.
1215
# XXX: At that point we need to consider the impact of double reads by
1216
# utilising components multiple times.
1217
if include_delta_closure:
1218
# XXX: get_content_maps performs its own index queries; allow state
1219
# to be passed in.
1220
text_map, _ = self._get_content_maps(present_keys,
1221
needed_from_fallback - absent_keys)
1222
for key in present_keys:
1223
yield FulltextContentFactory(key, global_map[key], None,
1224
''.join(text_map[key]))
1225
else:
1226
for source, keys in source_keys:
1227
if source is parent_maps[0]:
1228
# this KnitVersionedFiles
1229
records = [(key, positions[key][1]) for key in keys]
1230
for key, raw_data, sha1 in self._read_records_iter_raw(records):
1231
(record_details, index_memo, _) = positions[key]
1232
yield KnitContentFactory(key, global_map[key],
1233
record_details, sha1, raw_data, self._factory.annotated, None)
1234
else:
1235
vf = self._fallback_vfs[parent_maps.index(source) - 1]
1236
for record in vf.get_record_stream(keys, ordering,
1237
include_delta_closure):
1238
yield record
1239
1240
def get_sha1s(self, keys):
1241
"""See VersionedFiles.get_sha1s()."""
1242
missing = set(keys)
1243
record_map = self._get_record_map(missing, allow_missing=True)
1244
result = {}
1245
for key, details in record_map.iteritems():
1246
if key not in missing:
1247
continue
1248
# record entry 2 is the 'digest'.
1249
result[key] = details[2]
1250
missing.difference_update(set(result))
1251
for source in self._fallback_vfs:
1252
if not missing:
1253
break
1254
new_result = source.get_sha1s(missing)
1255
result.update(new_result)
1256
missing.difference_update(set(new_result))
1257
return result
1258
1259
def insert_record_stream(self, stream):
1260
"""Insert a record stream into this container.
1261
1262
:param stream: A stream of records to insert.
1263
:return: None
1264
:seealso VersionedFiles.get_record_stream:
1265
"""
1266
def get_adapter(adapter_key):
1267
try:
1268
return adapters[adapter_key]
1269
except KeyError:
1270
adapter_factory = adapter_registry.get(adapter_key)
1271
adapter = adapter_factory(self)
1272
adapters[adapter_key] = adapter
1273
return adapter
1274
if self._factory.annotated:
1275
# self is annotated, we need annotated knits to use directly.
1276
annotated = "annotated-"
1277
convertibles = []
1278
else:
1279
# self is not annotated, but we can strip annotations cheaply.
1280
annotated = ""
1281
convertibles = set(["knit-annotated-ft-gz"])
1282
if self._max_delta_chain:
1283
convertibles.add("knit-annotated-delta-gz")
1284
# The set of types we can cheaply adapt without needing basis texts.
1285
native_types = set()
1286
if self._max_delta_chain:
1287
native_types.add("knit-%sdelta-gz" % annotated)
1288
native_types.add("knit-%sft-gz" % annotated)
1289
knit_types = native_types.union(convertibles)
1290
adapters = {}
1291
# Buffer all index entries that we can't add immediately because their
1292
# basis parent is missing. We don't buffer all because generating
1293
# annotations may require access to some of the new records. However we
1294
# can't generate annotations from new deltas until their basis parent
1295
# is present anyway, so we get away with not needing an index that
1296
# includes the new keys.
1297
# key = basis_parent, value = index entry to add
1298
buffered_index_entries = {}
1299
for record in stream:
1300
parents = record.parents
1301
# Raise an error when a record is missing.
1302
if record.storage_kind == 'absent':
1303
raise RevisionNotPresent([record.key], self)
1304
if record.storage_kind in knit_types:
1305
if record.storage_kind not in native_types:
1306
try:
1307
adapter_key = (record.storage_kind, "knit-delta-gz")
1308
adapter = get_adapter(adapter_key)
1309
except KeyError:
1310
adapter_key = (record.storage_kind, "knit-ft-gz")
1311
adapter = get_adapter(adapter_key)
1312
bytes = adapter.get_bytes(
1313
record, record.get_bytes_as(record.storage_kind))
1314
else:
1315
bytes = record.get_bytes_as(record.storage_kind)
1316
options = [record._build_details[0]]
1317
if record._build_details[1]:
1318
options.append('no-eol')
1319
# Just blat it across.
1320
# Note: This does end up adding data on duplicate keys. As
1321
# modern repositories use atomic insertions this should not
1322
# lead to excessive growth in the event of interrupted fetches.
1323
# 'knit' repositories may suffer excessive growth, but as a
1324
# deprecated format this is tolerable. It can be fixed if
1325
# needed by in the kndx index support raising on a duplicate
1326
# add with identical parents and options.
1327
access_memo = self._access.add_raw_records(
1328
[(record.key, len(bytes))], bytes)[0]
1329
index_entry = (record.key, options, access_memo, parents)
1330
buffered = False
1331
if 'fulltext' not in options:
1332
basis_parent = parents[0]
1333
# Note that pack backed knits don't need to buffer here
1334
# because they buffer all writes to the transaction level,
1335
# but we don't expose that difference at the index level. If
1336
# the query here has sufficient cost to show up in
1337
# profiling we should do that.
1338
if basis_parent not in self.get_parent_map([basis_parent]):
1339
pending = buffered_index_entries.setdefault(
1340
basis_parent, [])
1341
pending.append(index_entry)
1342
buffered = True
1343
if not buffered:
1344
self._index.add_records([index_entry])
1345
elif record.storage_kind == 'fulltext':
1346
self.add_lines(record.key, parents,
1347
split_lines(record.get_bytes_as('fulltext')))
1348
else:
1349
adapter_key = record.storage_kind, 'fulltext'
1350
adapter = get_adapter(adapter_key)
1351
lines = split_lines(adapter.get_bytes(
1352
record, record.get_bytes_as(record.storage_kind)))
1353
try:
1354
self.add_lines(record.key, parents, lines)
1355
except errors.RevisionAlreadyPresent:
1356
pass
1357
# Add any records whose basis parent is now available.
1358
added_keys = [record.key]
1359
while added_keys:
1360
key = added_keys.pop(0)
1361
if key in buffered_index_entries:
1362
index_entries = buffered_index_entries[key]
1363
self._index.add_records(index_entries)
1364
added_keys.extend(
1365
[index_entry[0] for index_entry in index_entries])
1366
del buffered_index_entries[key]
1367
# If there were any deltas which had a missing basis parent, error.
1368
if buffered_index_entries:
1369
raise errors.RevisionNotPresent(buffered_index_entries.keys()[0],
1370
self)
1371
1372
def iter_lines_added_or_present_in_keys(self, keys, pb=None):
1373
"""Iterate over the lines in the versioned files from keys.
1374
1375
This may return lines from other keys. Each item the returned
1376
iterator yields is a tuple of a line and a text version that that line
1377
is present in (not introduced in).
1378
1379
Ordering of results is in whatever order is most suitable for the
1380
underlying storage format.
1381
1382
If a progress bar is supplied, it may be used to indicate progress.
1383
The caller is responsible for cleaning up progress bars (because this
1384
is an iterator).
1385
1386
NOTES:
1387
* Lines are normalised by the underlying store: they will all have \n
1388
terminators.
1389
* Lines are returned in arbitrary order.
1390
1391
:return: An iterator over (line, key).
1392
"""
1055
@staticmethod
1056
def _apply_delta(lines, delta):
1057
"""Apply delta to lines."""
1058
lines = list(lines)
1059
offset = 0
1060
for start, end, count, delta_lines in delta:
1061
lines[offset+start:offset+end] = delta_lines
1062
offset = offset + (start - end) + count
1063
return lines
1064
1065
def iter_lines_added_or_present_in_versions(self, version_ids=None,
1066
pb=None):
1067
"""See VersionedFile.iter_lines_added_or_present_in_versions()."""
1068
if version_ids is None:
1069
version_ids = self.versions()
1393
1070
if pb is None:
1394
1071
pb = progress.DummyProgress()
1395
keys = set(keys)
1396
total = len(keys)
1397
1072
# we don't care about inclusions, the caller cares.
1398
1073
# but we need to setup a list of records to visit.
1399
# we need key, position, length
1400
key_records = []
1401
build_details = self._index.get_build_details(keys)
1402
for key, details in build_details.iteritems():
1403
if key in keys:
1404
key_records.append((key, details[0]))
1405
keys.remove(key)
1406
records_iter = enumerate(self._read_records_iter(key_records))
1407
for (key_idx, (key, data, sha_value)) in records_iter:
1408
pb.update('Walking content.', key_idx, total)
1409
compression_parent = build_details[key][1]
1410
if compression_parent is None:
1411
# fulltext
1412
line_iterator = self._factory.get_fulltext_content(data)
1074
# we need version_id, position, length
1075
version_id_records = []
1076
requested_versions = set(version_ids)
1077
# filter for available versions
1078
for version_id in requested_versions:
1079
if not self.has_version(version_id):
1080
raise RevisionNotPresent(version_id, self.filename)
1081
# get a in-component-order queue:
1082
for version_id in self.versions():
1083
if version_id in requested_versions:
1084
index_memo = self._index.get_position(version_id)
1085
version_id_records.append((version_id, index_memo))
1086
1087
total = len(version_id_records)
1088
for version_idx, (version_id, data, sha_value) in \
1089
enumerate(self._data.read_records_iter(version_id_records)):
1090
pb.update('Walking content.', version_idx, total)
1091
method = self._index.get_method(version_id)
1092
1093
assert method in ('fulltext', 'line-delta')
1094
if method == 'fulltext':
1095
line_iterator = self.factory.get_fulltext_content(data)
1413
1096
else:
1414
# Delta
1415
line_iterator = self._factory.get_linedelta_content(data)
1416
# XXX: It might be more efficient to yield (key,
1417
# line_iterator) in the future. However for now, this is a simpler
1418
# change to integrate into the rest of the codebase. RBC 20071110
1097
line_iterator = self.factory.get_linedelta_content(data)
1419
1098
for line in line_iterator:
1420
yield line, key
1421
for source in self._fallback_vfs:
1422
if not keys:
1423
break
1424
source_keys = set()
1425
for line, key in source.iter_lines_added_or_present_in_keys(keys):
1426
source_keys.add(key)
1427
yield line, key
1428
keys.difference_update(source_keys)
1429
if keys:
1430
raise RevisionNotPresent(keys, self.filename)
1099
yield line
1100
1431
1101
pb.update('Walking content.', total, total)
1432
1433
def _make_line_delta(self, delta_seq, new_content):
1434
"""Generate a line delta from delta_seq and new_content."""
1435
diff_hunks = []
1436
for op in delta_seq.get_opcodes():
1437
if op[0] == 'equal':
1438
continue
1439
diff_hunks.append((op[1], op[2], op[4]-op[3], new_content._lines[op[3]:op[4]]))
1440
return diff_hunks
1441
1442
def _merge_annotations(self, content, parents, parent_texts={},
1443
delta=None, annotated=None,
1444
left_matching_blocks=None):
1445
"""Merge annotations for content and generate deltas.
1446
1447
This is done by comparing the annotations based on changes to the text
1448
and generating a delta on the resulting full texts. If annotations are
1449
not being created then a simple delta is created.
1450
"""
1451
if left_matching_blocks is not None:
1452
delta_seq = diff._PrematchedMatcher(left_matching_blocks)
1453
else:
1454
delta_seq = None
1455
if annotated:
1456
for parent_key in parents:
1457
merge_content = self._get_content(parent_key, parent_texts)
1458
if (parent_key == parents[0] and delta_seq is not None):
1459
seq = delta_seq
1460
else:
1461
seq = patiencediff.PatienceSequenceMatcher(
1462
None, merge_content.text(), content.text())
1463
for i, j, n in seq.get_matching_blocks():
1464
if n == 0:
1465
continue
1466
# this copies (origin, text) pairs across to the new
1467
# content for any line that matches the last-checked
1468
# parent.
1469
content._lines[j:j+n] = merge_content._lines[i:i+n]
1470
# XXX: Robert says the following block is a workaround for a
1471
# now-fixed bug and it can probably be deleted. -- mbp 20080618
1472
if content._lines and content._lines[-1][1][-1] != '\n':
1473
# The copied annotation was from a line without a trailing EOL,
1474
# reinstate one for the content object, to ensure correct
1475
# serialization.
1476
line = content._lines[-1][1] + '\n'
1477
content._lines[-1] = (content._lines[-1][0], line)
1478
if delta:
1479
if delta_seq is None:
1480
reference_content = self._get_content(parents[0], parent_texts)
1481
new_texts = content.text()
1482
old_texts = reference_content.text()
1483
delta_seq = patiencediff.PatienceSequenceMatcher(
1484
None, old_texts, new_texts)
1485
return self._make_line_delta(delta_seq, content)
1486
1487
def _parse_record(self, version_id, data):
1488
"""Parse an original format knit record.
1489
1490
These have the last element of the key only present in the stored data.
1491
"""
1492
rec, record_contents = self._parse_record_unchecked(data)
1493
self._check_header_version(rec, version_id)
1494
return record_contents, rec[3]
1495
1496
def _parse_record_header(self, key, raw_data):
1497
"""Parse a record header for consistency.
1498
1499
:return: the header and the decompressor stream.
1500
as (stream, header_record)
1501
"""
1502
df = tuned_gzip.GzipFile(mode='rb', fileobj=StringIO(raw_data))
1503
try:
1504
# Current serialise
1505
rec = self._check_header(key, df.readline())
1506
except Exception, e:
1507
raise KnitCorrupt(self,
1508
"While reading {%s} got %s(%s)"
1509
% (key, e.__class__.__name__, str(e)))
1510
return df, rec
1511
1512
def _parse_record_unchecked(self, data):
1513
# profiling notes:
1514
# 4168 calls in 2880 217 internal
1515
# 4168 calls to _parse_record_header in 2121
1516
# 4168 calls to readlines in 330
1517
df = tuned_gzip.GzipFile(mode='rb', fileobj=StringIO(data))
1518
try:
1519
record_contents = df.readlines()
1520
except Exception, e:
1521
raise KnitCorrupt(self, "Corrupt compressed record %r, got %s(%s)" %
1522
(data, e.__class__.__name__, str(e)))
1523
header = record_contents.pop(0)
1524
rec = self._split_header(header)
1525
last_line = record_contents.pop()
1526
if len(record_contents) != int(rec[2]):
1527
raise KnitCorrupt(self,
1528
'incorrect number of lines %s != %s'
1529
' for version {%s} %s'
1530
% (len(record_contents), int(rec[2]),
1531
rec[1], record_contents))
1532
if last_line != 'end %s\n' % rec[1]:
1533
raise KnitCorrupt(self,
1534
'unexpected version end line %r, wanted %r'
1535
% (last_line, rec[1]))
1536
df.close()
1537
return rec, record_contents
1538
1539
def _read_records_iter(self, records):
1540
"""Read text records from data file and yield result.
1541
1542
The result will be returned in whatever is the fastest to read.
1543
Not by the order requested. Also, multiple requests for the same
1544
record will only yield 1 response.
1545
:param records: A list of (key, access_memo) entries
1546
:return: Yields (key, contents, digest) in the order
1547
read, not the order requested
1548
"""
1549
if not records:
1550
return
1551
1552
# XXX: This smells wrong, IO may not be getting ordered right.
1553
needed_records = sorted(set(records), key=operator.itemgetter(1))
1554
if not needed_records:
1555
return
1556
1557
# The transport optimizes the fetching as well
1558
# (ie, reads continuous ranges.)
1559
raw_data = self._access.get_raw_records(
1560
[index_memo for key, index_memo in needed_records])
1561
1562
for (key, index_memo), data in \
1563
izip(iter(needed_records), raw_data):
1564
content, digest = self._parse_record(key[-1], data)
1565
yield key, content, digest
1566
1567
def _read_records_iter_raw(self, records):
1568
"""Read text records from data file and yield raw data.
1569
1570
This unpacks enough of the text record to validate the id is
1571
as expected but thats all.
1572
1573
Each item the iterator yields is (key, bytes, sha1_of_full_text).
1574
"""
1575
# setup an iterator of the external records:
1576
# uses readv so nice and fast we hope.
1577
if len(records):
1578
# grab the disk data needed.
1579
needed_offsets = [index_memo for key, index_memo
1580
in records]
1581
raw_records = self._access.get_raw_records(needed_offsets)
1582
1583
for key, index_memo in records:
1584
data = raw_records.next()
1585
# validate the header (note that we can only use the suffix in
1586
# current knit records).
1587
df, rec = self._parse_record_header(key, data)
1588
df.close()
1589
yield key, data, rec[3]
1590
1591
def _record_to_data(self, key, digest, lines, dense_lines=None):
1592
"""Convert key, digest, lines into a raw data block.
1593
1594
:param key: The key of the record. Currently keys are always serialised
1595
using just the trailing component.
1596
:param dense_lines: The bytes of lines but in a denser form. For
1597
instance, if lines is a list of 1000 bytestrings each ending in \n,
1598
dense_lines may be a list with one line in it, containing all the
1599
1000's lines and their \n's. Using dense_lines if it is already
1600
known is a win because the string join to create bytes in this
1601
function spends less time resizing the final string.
1602
:return: (len, a StringIO instance with the raw data ready to read.)
1603
"""
1604
# Note: using a string copy here increases memory pressure with e.g.
1605
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
1606
# when doing the initial commit of a mozilla tree. RBC 20070921
1607
bytes = ''.join(chain(
1608
["version %s %d %s\n" % (key[-1],
1609
len(lines),
1610
digest)],
1611
dense_lines or lines,
1612
["end %s\n" % key[-1]]))
1613
if type(bytes) != str:
1614
raise AssertionError(
1615
'data must be plain bytes was %s' % type(bytes))
1616
if lines and lines[-1][-1] != '\n':
1617
raise ValueError('corrupt lines value %r' % lines)
1618
compressed_bytes = tuned_gzip.bytes_to_gzip(bytes)
1619
return len(compressed_bytes), compressed_bytes
1620
1621
def _split_header(self, line):
1622
rec = line.split()
1623
if len(rec) != 4:
1624
raise KnitCorrupt(self,
1625
'unexpected number of elements in record header')
1626
return rec
1627
1628
def keys(self):
1629
"""See VersionedFiles.keys."""
1630
if 'evil' in debug.debug_flags:
1631
trace.mutter_callsite(2, "keys scales with size of history")
1632
sources = [self._index] + self._fallback_vfs
1633
result = set()
1634
for source in sources:
1635
result.update(source.keys())
1636
return result
1637
1638
1639
1640
class _KndxIndex(object):
1641
"""Manages knit index files
1642
1643
The index is kept in memory and read on startup, to enable
1102
1103
def iter_parents(self, version_ids):
1104
"""Iterate through the parents for many version ids.
1105
1106
:param version_ids: An iterable yielding version_ids.
1107
:return: An iterator that yields (version_id, parents). Requested
1108
version_ids not present in the versioned file are simply skipped.
1109
The order is undefined, allowing for different optimisations in
1110
the underlying implementation.
1111
"""
1112
return self._index.iter_parents(version_ids)
1113
1114
def num_versions(self):
1115
"""See VersionedFile.num_versions()."""
1116
return self._index.num_versions()
1117
1118
__len__ = num_versions
1119
1120
def annotate_iter(self, version_id):
1121
"""See VersionedFile.annotate_iter."""
1122
return self.factory.annotate_iter(self, version_id)
1123
1124
def get_parents(self, version_id):
1125
"""See VersionedFile.get_parents."""
1126
# perf notes:
1127
# optimism counts!
1128
# 52554 calls in 1264 872 internal down from 3674
1129
try:
1130
return self._index.get_parents(version_id)
1131
except KeyError:
1132
raise RevisionNotPresent(version_id, self.filename)
1133
1134
def get_parents_with_ghosts(self, version_id):
1135
"""See VersionedFile.get_parents."""
1136
try:
1137
return self._index.get_parents_with_ghosts(version_id)
1138
except KeyError:
1139
raise RevisionNotPresent(version_id, self.filename)
1140
1141
def get_ancestry(self, versions, topo_sorted=True):
1142
"""See VersionedFile.get_ancestry."""
1143
if isinstance(versions, basestring):
1144
versions = [versions]
1145
if not versions:
1146
return []
1147
return self._index.get_ancestry(versions, topo_sorted)
1148
1149
def get_ancestry_with_ghosts(self, versions):
1150
"""See VersionedFile.get_ancestry_with_ghosts."""
1151
if isinstance(versions, basestring):
1152
versions = [versions]
1153
if not versions:
1154
return []
1155
return self._index.get_ancestry_with_ghosts(versions)
1156
1157
def plan_merge(self, ver_a, ver_b):
1158
"""See VersionedFile.plan_merge."""
1159
ancestors_b = set(self.get_ancestry(ver_b, topo_sorted=False))
1160
ancestors_a = set(self.get_ancestry(ver_a, topo_sorted=False))
1161
annotated_a = self.annotate(ver_a)
1162
annotated_b = self.annotate(ver_b)
1163
return merge._plan_annotate_merge(annotated_a, annotated_b,
1164
ancestors_a, ancestors_b)
1165
1166
1167
class _KnitComponentFile(object):
1168
"""One of the files used to implement a knit database"""
1169
1170
def __init__(self, transport, filename, mode, file_mode=None,
1171
create_parent_dir=False, dir_mode=None):
1172
self._transport = transport
1173
self._filename = filename
1174
self._mode = mode
1175
self._file_mode = file_mode
1176
self._dir_mode = dir_mode
1177
self._create_parent_dir = create_parent_dir
1178
self._need_to_create = False
1179
1180
def _full_path(self):
1181
"""Return the full path to this file."""
1182
return self._transport.base + self._filename
1183
1184
def check_header(self, fp):
1185
line = fp.readline()
1186
if line == '':
1187
# An empty file can actually be treated as though the file doesn't
1188
# exist yet.
1189
raise errors.NoSuchFile(self._full_path())
1190
if line != self.HEADER:
1191
raise KnitHeaderError(badline=line,
1192
filename=self._transport.abspath(self._filename))
1193
1194
def __repr__(self):
1195
return '%s(%s)' % (self.__class__.__name__, self._filename)
1196
1197
1198
class _KnitIndex(_KnitComponentFile):
1199
"""Manages knit index file.
1200
1201
The index is already kept in memory and read on startup, to enable
1644
1202
fast lookups of revision information. The cursor of the index
1645
1203
file is always pointing to the end, making it easy to append
1646
1204
entries.
1688
1246
to ensure that records always start on new lines even if the last write was
1689
1247
interrupted. As a result its normal for the last line in the index to be
1690
1248
missing a trailing newline. One can be added with no harmful effects.
1691
1692
:ivar _kndx_cache: dict from prefix to the old state of KnitIndex objects,
1693
where prefix is e.g. the (fileid,) for .texts instances or () for
1694
constant-mapped things like .revisions, and the old state is
1695
tuple(history_vector, cache_dict). This is used to prevent having an
1696
ABI change with the C extension that reads .kndx files.
1697
1249
"""
1698
1250
1699
1251
HEADER = "# bzr knit index 8\n"
1700
1252
1701
def __init__(self, transport, mapper, get_scope, allow_writes, is_locked):
1702
"""Create a _KndxIndex on transport using mapper."""
1703
self._transport = transport
1704
self._mapper = mapper
1705
self._get_scope = get_scope
1706
self._allow_writes = allow_writes
1707
self._is_locked = is_locked
1708
self._reset_cache()
1709
self.has_graph = True
1710
1711
def add_records(self, records, random_id=False):
1712
"""Add multiple records to the index.
1713
1714
:param records: a list of tuples:
1715
(key, options, access_memo, parents).
1716
:param random_id: If True the ids being added were randomly generated
1717
and no check for existence will be performed.
1718
"""
1719
paths = {}
1720
for record in records:
1721
key = record[0]
1722
prefix = key[:-1]
1723
path = self._mapper.map(key) + '.kndx'
1724
path_keys = paths.setdefault(path, (prefix, []))
1725
path_keys[1].append(record)
1726
for path in sorted(paths):
1727
prefix, path_keys = paths[path]
1728
self._load_prefixes([prefix])
1729
lines = []
1730
orig_history = self._kndx_cache[prefix][1][:]
1731
orig_cache = self._kndx_cache[prefix][0].copy()
1732
1733
try:
1734
for key, options, (_, pos, size), parents in path_keys:
1735
if parents is None:
1736
# kndx indices cannot be parentless.
1737
parents = ()
1738
line = "\n%s %s %s %s %s :" % (
1739
key[-1], ','.join(options), pos, size,
1740
self._dictionary_compress(parents))
1741
if type(line) != str:
1742
raise AssertionError(
1743
'data must be utf8 was %s' % type(line))
1744
lines.append(line)
1745
self._cache_key(key, options, pos, size, parents)
1746
if len(orig_history):
1747
self._transport.append_bytes(path, ''.join(lines))
1748
else:
1749
self._init_index(path, lines)
1750
except:
1751
# If any problems happen, restore the original values and re-raise
1752
self._kndx_cache[prefix] = (orig_cache, orig_history)
1753
raise
1754
1755
def _cache_key(self, key, options, pos, size, parent_keys):
1253
# speed of knit parsing went from 280 ms to 280 ms with slots addition.
1254
# __slots__ = ['_cache', '_history', '_transport', '_filename']
1255
1256
def _cache_version(self, version_id, options, pos, size, parents):
1756
1257
"""Cache a version record in the history array and index cache.
1757
1258
1758
1259
This is inlined into _load_data for performance. KEEP IN SYNC.
1759
1260
(It saves 60ms, 25% of the __init__ overhead on local 4000 record
1760
1261
indexes).
1761
1262
"""
1762
prefix = key[:-1]
1763
version_id = key[-1]
1764
# last-element only for compatibilty with the C load_data.
1765
parents = tuple(parent[-1] for parent in parent_keys)
1766
for parent in parent_keys:
1767
if parent[:-1] != prefix:
1768
raise ValueError("mismatched prefixes for %r, %r" % (
1769
key, parent_keys))
1770
cache, history = self._kndx_cache[prefix]
1771
1263
# only want the _history index to reference the 1st index entry
1772
1264
# for version_id
1773
if version_id not in cache:
1774
index = len(history)
1775
history.append(version_id)
1265
if version_id not in self._cache:
1266
index = len(self._history)
1267
self._history.append(version_id)
1776
1268
else:
1777
index = cache[version_id][5]
1778
cache[version_id] = (version_id,
1269
index = self._cache[version_id][5]
1270
self._cache[version_id] = (version_id,
1779
1271
options,
1780
1272
pos,
1781
1273
size,
1782
1274
parents,
1783
1275
index)
1784
1276
1785
def check_header(self, fp):
1786
line = fp.readline()
1787
if line == '':
1788
# An empty file can actually be treated as though the file doesn't
1789
# exist yet.
1790
raise errors.NoSuchFile(self)
1791
if line != self.HEADER:
1792
raise KnitHeaderError(badline=line, filename=self)
1793
1794
def _check_read(self):
1795
if not self._is_locked():
1796
raise errors.ObjectNotLocked(self)
1797
if self._get_scope() != self._scope:
1798
self._reset_cache()
1799
1800
def _check_write_ok(self):
1801
"""Assert if not writes are permitted."""
1802
if not self._is_locked():
1803
raise errors.ObjectNotLocked(self)
1804
if self._get_scope() != self._scope:
1805
self._reset_cache()
1806
if self._mode != 'w':
1807
raise errors.ReadOnlyObjectDirtiedError(self)
1808
1809
def get_build_details(self, keys):
1810
"""Get the method, index_memo and compression parent for keys.
1811
1812
Ghosts are omitted from the result.
1813
1814
:param keys: An iterable of keys.
1815
:return: A dict of key:(index_memo, compression_parent, parents,
1816
record_details).
1817
index_memo
1818
opaque structure to pass to read_records to extract the raw
1819
data
1820
compression_parent
1821
Content that this record is built upon, may be None
1822
parents
1823
Logical parents of this node
1824
record_details
1825
extra information about the content which needs to be passed to
1826
Factory.parse_record
1827
"""
1828
prefixes = self._partition_keys(keys)
1829
parent_map = self.get_parent_map(keys)
1830
result = {}
1831
for key in keys:
1832
if key not in parent_map:
1833
continue # Ghost
1834
method = self.get_method(key)
1835
parents = parent_map[key]
1836
if method == 'fulltext':
1837
compression_parent = None
1838
else:
1839
compression_parent = parents[0]
1840
noeol = 'no-eol' in self.get_options(key)
1841
index_memo = self.get_position(key)
1842
result[key] = (index_memo, compression_parent,
1843
parents, (method, noeol))
1844
return result
1845
1846
def get_method(self, key):
1847
"""Return compression method of specified key."""
1848
options = self.get_options(key)
1849
if 'fulltext' in options:
1850
return 'fulltext'
1851
elif 'line-delta' in options:
1852
return 'line-delta'
1853
else:
1854
raise errors.KnitIndexUnknownMethod(self, options)
1855
1856
def get_options(self, key):
1857
"""Return a list representing options.
1858
1859
e.g. ['foo', 'bar']
1860
"""
1861
prefix, suffix = self._split_key(key)
1862
self._load_prefixes([prefix])
1277
def __init__(self, transport, filename, mode, create=False, file_mode=None,
1278
create_parent_dir=False, delay_create=False, dir_mode=None):
1279
_KnitComponentFile.__init__(self, transport, filename, mode,
1280
file_mode=file_mode,
1281
create_parent_dir=create_parent_dir,
1282
dir_mode=dir_mode)
1283
self._cache = {}
1284
# position in _history is the 'official' index for a revision
1285
# but the values may have come from a newer entry.
1286
# so - wc -l of a knit index is != the number of unique names
1287
# in the knit.
1288
self._history = []
1863
1289
try:
1864
return self._kndx_cache[prefix][0][suffix][1]
1865
except KeyError:
1866
raise RevisionNotPresent(key, self)
1867
1868
def get_parent_map(self, keys):
1869
"""Get a map of the parents of keys.
1870
1871
:param keys: The keys to look up parents for.
1872
:return: A mapping from keys to parents. Absent keys are absent from
1873
the mapping.
1290
fp = self._transport.get(self._filename)
1291
try:
1292
# _load_data may raise NoSuchFile if the target knit is
1293
# completely empty.
1294
_load_data(self, fp)
1295
finally:
1296
fp.close()
1297
except NoSuchFile:
1298
if mode != 'w' or not create:
1299
raise
1300
elif delay_create:
1301
self._need_to_create = True
1302
else:
1303
self._transport.put_bytes_non_atomic(
1304
self._filename, self.HEADER, mode=self._file_mode)
1305
1306
def get_graph(self):
1307
"""Return a list of the node:parents lists from this knit index."""
1308
return [(vid, idx[4]) for vid, idx in self._cache.iteritems()]
1309
1310
def get_ancestry(self, versions, topo_sorted=True):
1311
"""See VersionedFile.get_ancestry."""
1312
# get a graph of all the mentioned versions:
1313
graph = {}
1314
pending = set(versions)
1315
cache = self._cache
1316
while pending:
1317
version = pending.pop()
1318
# trim ghosts
1319
try:
1320
parents = [p for p in cache[version][4] if p in cache]
1321
except KeyError:
1322
raise RevisionNotPresent(version, self._filename)
1323
# if not completed and not a ghost
1324
pending.update([p for p in parents if p not in graph])
1325
graph[version] = parents
1326
if not topo_sorted:
1327
return graph.keys()
1328
return topo_sort(graph.items())
1329
1330
def get_ancestry_with_ghosts(self, versions):
1331
"""See VersionedFile.get_ancestry_with_ghosts."""
1332
# get a graph of all the mentioned versions:
1333
self.check_versions_present(versions)
1334
cache = self._cache
1335
graph = {}
1336
pending = set(versions)
1337
while pending:
1338
version = pending.pop()
1339
try:
1340
parents = cache[version][4]
1341
except KeyError:
1342
# ghost, fake it
1343
graph[version] = []
1344
else:
1345
# if not completed
1346
pending.update([p for p in parents if p not in graph])
1347
graph[version] = parents
1348
return topo_sort(graph.items())
1349
1350
def iter_parents(self, version_ids):
1351
"""Iterate through the parents for many version ids.
1352
1353
:param version_ids: An iterable yielding version_ids.
1354
:return: An iterator that yields (version_id, parents). Requested
1355
version_ids not present in the versioned file are simply skipped.
1356
The order is undefined, allowing for different optimisations in
1357
the underlying implementation.
1874
1358
"""
1875
# Parse what we need to up front, this potentially trades off I/O
1876
# locality (.kndx and .knit in the same block group for the same file
1877
# id) for less checking in inner loops.
1878
prefixes = set(key[:-1] for key in keys)
1879
self._load_prefixes(prefixes)
1880
result = {}
1881
for key in keys:
1882
prefix = key[:-1]
1359
for version_id in version_ids:
1883
1360
try:
1884
suffix_parents = self._kndx_cache[prefix][0][key[-1]][4]
1361
yield version_id, tuple(self.get_parents(version_id))
1885
1362
except KeyError:
1886
1363
pass
1887
else:
1888
result[key] = tuple(prefix + (suffix,) for
1889
suffix in suffix_parents)
1890
return result
1891
1892
def get_position(self, key):
1893
"""Return details needed to access the version.
1894
1895
:return: a tuple (key, data position, size) to hand to the access
1896
logic to get the record.
1897
"""
1898
prefix, suffix = self._split_key(key)
1899
self._load_prefixes([prefix])
1900
entry = self._kndx_cache[prefix][0][suffix]
1901
return key, entry[2], entry[3]
1902
1903
def _init_index(self, path, extra_lines=[]):
1904
"""Initialize an index."""
1905
sio = StringIO()
1906
sio.write(self.HEADER)
1907
sio.writelines(extra_lines)
1908
sio.seek(0)
1909
self._transport.put_file_non_atomic(path, sio,
1910
create_parent_dir=True)
1911
# self._create_parent_dir)
1912
# mode=self._file_mode,
1913
# dir_mode=self._dir_mode)
1914
1915
def keys(self):
1916
"""Get all the keys in the collection.
1917
1918
The keys are not ordered.
1919
"""
1920
result = set()
1921
# Identify all key prefixes.
1922
# XXX: A bit hacky, needs polish.
1923
if type(self._mapper) == ConstantMapper:
1924
prefixes = [()]
1925
else:
1926
relpaths = set()
1927
for quoted_relpath in self._transport.iter_files_recursive():
1928
path, ext = os.path.splitext(quoted_relpath)
1929
relpaths.add(path)
1930
prefixes = [self._mapper.unmap(path) for path in relpaths]
1931
self._load_prefixes(prefixes)
1932
for prefix in prefixes:
1933
for suffix in self._kndx_cache[prefix][1]:
1934
result.add(prefix + (suffix,))
1935
return result
1936
1937
def _load_prefixes(self, prefixes):
1938
"""Load the indices for prefixes."""
1939
self._check_read()
1940
for prefix in prefixes:
1941
if prefix not in self._kndx_cache:
1942
# the load_data interface writes to these variables.
1943
self._cache = {}
1944
self._history = []
1945
self._filename = prefix
1946
try:
1947
path = self._mapper.map(prefix) + '.kndx'
1948
fp = self._transport.get(path)
1949
try:
1950
# _load_data may raise NoSuchFile if the target knit is
1951
# completely empty.
1952
_load_data(self, fp)
1953
finally:
1954
fp.close()
1955
self._kndx_cache[prefix] = (self._cache, self._history)
1956
del self._cache
1957
del self._filename
1958
del self._history
1959
except NoSuchFile:
1960
self._kndx_cache[prefix] = ({}, [])
1961
if type(self._mapper) == ConstantMapper:
1962
# preserve behaviour for revisions.kndx etc.
1963
self._init_index(path)
1964
del self._cache
1965
del self._filename
1966
del self._history
1967
1968
def _partition_keys(self, keys):
1969
"""Turn keys into a dict of prefix:suffix_list."""
1970
result = {}
1971
for key in keys:
1972
prefix_keys = result.setdefault(key[:-1], [])
1973
prefix_keys.append(key[-1])
1974
return result
1975
1976
def _dictionary_compress(self, keys):
1977
"""Dictionary compress keys.
1978
1979
:param keys: The keys to generate references to.
1980
:return: A string representation of keys. keys which are present are
1981
dictionary compressed, and others are emitted as fulltext with a
1982
'.' prefix.
1983
"""
1984
if not keys:
1985
return ''
1364
1365
def num_versions(self):
1366
return len(self._history)
1367
1368
__len__ = num_versions
1369
1370
def get_versions(self):
1371
"""Get all the versions in the file. not topologically sorted."""
1372
return self._history
1373
1374
def _version_list_to_index(self, versions):
1986
1375
result_list = []
1987
prefix = keys[0][:-1]
1988
cache = self._kndx_cache[prefix][0]
1989
for key in keys:
1990
if key[:-1] != prefix:
1991
# kndx indices cannot refer across partitioned storage.
1992
raise ValueError("mismatched prefixes for %r" % keys)
1993
if key[-1] in cache:
1376
cache = self._cache
1377
for version in versions:
1378
if version in cache:
1994
1379
# -- inlined lookup() --
1995
result_list.append(str(cache[key[-1]][5]))
1380
result_list.append(str(cache[version][5]))
1996
1381
# -- end lookup () --
1997
1382
else:
1998
result_list.append('.' + key[-1])
1383
result_list.append('.' + version)
1999
1384
return ' '.join(result_list)
2000
1385
2001
def _reset_cache(self):
2002
# Possibly this should be a LRU cache. A dictionary from key_prefix to
2003
# (cache_dict, history_vector) for parsed kndx files.
2004
self._kndx_cache = {}
2005
self._scope = self._get_scope()
2006
allow_writes = self._allow_writes()
2007
if allow_writes:
2008
self._mode = 'w'
1386
def add_version(self, version_id, options, index_memo, parents):
1387
"""Add a version record to the index."""
1388
self.add_versions(((version_id, options, index_memo, parents),))
1389
1390
def add_versions(self, versions, random_id=False):
1391
"""Add multiple versions to the index.
1392
1393
:param versions: a list of tuples:
1394
(version_id, options, pos, size, parents).
1395
:param random_id: If True the ids being added were randomly generated
1396
and no check for existence will be performed.
1397
"""
1398
lines = []
1399
orig_history = self._history[:]
1400
orig_cache = self._cache.copy()
1401
1402
try:
1403
for version_id, options, (index, pos, size), parents in versions:
1404
line = "\n%s %s %s %s %s :" % (version_id,
1405
','.join(options),
1406
pos,
1407
size,
1408
self._version_list_to_index(parents))
1409
assert isinstance(line, str), \
1410
'content must be utf-8 encoded: %r' % (line,)
1411
lines.append(line)
1412
self._cache_version(version_id, options, pos, size, parents)
1413
if not self._need_to_create:
1414
self._transport.append_bytes(self._filename, ''.join(lines))
1415
else:
1416
sio = StringIO()
1417
sio.write(self.HEADER)
1418
sio.writelines(lines)
1419
sio.seek(0)
1420
self._transport.put_file_non_atomic(self._filename, sio,
1421
create_parent_dir=self._create_parent_dir,
1422
mode=self._file_mode,
1423
dir_mode=self._dir_mode)
1424
self._need_to_create = False
1425
except:
1426
# If any problems happen, restore the original values and re-raise
1427
self._history = orig_history
1428
self._cache = orig_cache
1429
raise
1430
1431
def has_version(self, version_id):
1432
"""True if the version is in the index."""
1433
return version_id in self._cache
1434
1435
def get_position(self, version_id):
1436
"""Return details needed to access the version.
1437
1438
.kndx indices do not support split-out data, so return None for the
1439
index field.
1440
1441
:return: a tuple (None, data position, size) to hand to the access
1442
logic to get the record.
1443
"""
1444
entry = self._cache[version_id]
1445
return None, entry[2], entry[3]
1446
1447
def get_method(self, version_id):
1448
"""Return compression method of specified version."""
1449
try:
1450
options = self._cache[version_id][1]
1451
except KeyError:
1452
raise RevisionNotPresent(version_id, self._filename)
1453
if 'fulltext' in options:
1454
return 'fulltext'
2009
1455
else:
2010
self._mode = 'r'
2011
2012
def _split_key(self, key):
2013
"""Split key into a prefix and suffix."""
2014
return key[:-1], key[-1]
2015
2016
2017
class _KnitGraphIndex(object):
2018
"""A KnitVersionedFiles index layered on GraphIndex."""
2019
2020
def __init__(self, graph_index, is_locked, deltas=False, parents=True,
2021
add_callback=None):
1456
if 'line-delta' not in options:
1457
raise errors.KnitIndexUnknownMethod(self._full_path(), options)
1458
return 'line-delta'
1459
1460
def get_options(self, version_id):
1461
"""Return a string represention options.
1462
1463
e.g. foo,bar
1464
"""
1465
return self._cache[version_id][1]
1466
1467
def get_parents(self, version_id):
1468
"""Return parents of specified version ignoring ghosts."""
1469
return [parent for parent in self._cache[version_id][4]
1470
if parent in self._cache]
1471
1472
def get_parents_with_ghosts(self, version_id):
1473
"""Return parents of specified version with ghosts."""
1474
return self._cache[version_id][4]
1475
1476
def check_versions_present(self, version_ids):
1477
"""Check that all specified versions are present."""
1478
cache = self._cache
1479
for version_id in version_ids:
1480
if version_id not in cache:
1481
raise RevisionNotPresent(version_id, self._filename)
1482
1483
1484
class KnitGraphIndex(object):
1485
"""A knit index that builds on GraphIndex."""
1486
1487
def __init__(self, graph_index, deltas=False, parents=True, add_callback=None):
2022
1488
"""Construct a KnitGraphIndex on a graph_index.
2023
1489
2024
1490
:param graph_index: An implementation of bzrlib.index.GraphIndex.
2025
:param is_locked: A callback to check whether the object should answer
2026
queries.
2027
1491
:param deltas: Allow delta-compressed records.
2028
:param parents: If True, record knits parents, if not do not record
2029
parents.
2030
1492
:param add_callback: If not None, allow additions to the index and call
2031
1493
this callback with a list of added GraphIndex nodes:
2032
1494
[(node, value, node_refs), ...]
2033
:param is_locked: A callback, returns True if the index is locked and
2034
thus usable.
1495
:param parents: If True, record knits parents, if not do not record
1496
parents.
2035
1497
"""
2036
self._add_callback = add_callback
2037
1498
self._graph_index = graph_index
2038
1499
self._deltas = deltas
1500
self._add_callback = add_callback
2039
1501
self._parents = parents
2040
1502
if deltas and not parents:
2041
# XXX: TODO: Delta tree and parent graph should be conceptually
2042
# separate.
2043
1503
raise KnitCorrupt(self, "Cannot do delta compression without "
2044
1504
"parent tracking.")
2045
self.has_graph = parents
2046
self._is_locked = is_locked
2047
2048
def __repr__(self):
2049
return "%s(%r)" % (self.__class__.__name__, self._graph_index)
2050
2051
def add_records(self, records, random_id=False):
2052
"""Add multiple records to the index.
1505
1506
def _get_entries(self, keys, check_present=False):
1507
"""Get the entries for keys.
1508
1509
:param keys: An iterable of index keys, - 1-tuples.
1510
"""
1511
keys = set(keys)
1512
found_keys = set()
1513
if self._parents:
1514
for node in self._graph_index.iter_entries(keys):
1515
yield node
1516
found_keys.add(node[1])
1517
else:
1518
# adapt parentless index to the rest of the code.
1519
for node in self._graph_index.iter_entries(keys):
1520
yield node[0], node[1], node[2], ()
1521
found_keys.add(node[1])
1522
if check_present:
1523
missing_keys = keys.difference(found_keys)
1524
if missing_keys:
1525
raise RevisionNotPresent(missing_keys.pop(), self)
1526
1527
def _present_keys(self, version_ids):
1528
return set([
1529
node[1] for node in self._get_entries(version_ids)])
1530
1531
def _parentless_ancestry(self, versions):
1532
"""Honour the get_ancestry API for parentless knit indices."""
1533
wanted_keys = self._version_ids_to_keys(versions)
1534
present_keys = self._present_keys(wanted_keys)
1535
missing = set(wanted_keys).difference(present_keys)
1536
if missing:
1537
raise RevisionNotPresent(missing.pop(), self)
1538
return list(self._keys_to_version_ids(present_keys))
1539
1540
def get_ancestry(self, versions, topo_sorted=True):
1541
"""See VersionedFile.get_ancestry."""
1542
if not self._parents:
1543
return self._parentless_ancestry(versions)
1544
# XXX: This will do len(history) index calls - perhaps
1545
# it should be altered to be a index core feature?
1546
# get a graph of all the mentioned versions:
1547
graph = {}
1548
ghosts = set()
1549
versions = self._version_ids_to_keys(versions)
1550
pending = set(versions)
1551
while pending:
1552
# get all pending nodes
1553
this_iteration = pending
1554
new_nodes = self._get_entries(this_iteration)
1555
found = set()
1556
pending = set()
1557
for (index, key, value, node_refs) in new_nodes:
1558
# dont ask for ghosties - otherwise
1559
# we we can end up looping with pending
1560
# being entirely ghosted.
1561
graph[key] = [parent for parent in node_refs[0]
1562
if parent not in ghosts]
1563
# queue parents
1564
for parent in graph[key]:
1565
# dont examine known nodes again
1566
if parent in graph:
1567
continue
1568
pending.add(parent)
1569
found.add(key)
1570
ghosts.update(this_iteration.difference(found))
1571
if versions.difference(graph):
1572
raise RevisionNotPresent(versions.difference(graph).pop(), self)
1573
if topo_sorted:
1574
result_keys = topo_sort(graph.items())
1575
else:
1576
result_keys = graph.iterkeys()
1577
return [key[0] for key in result_keys]
1578
1579
def get_ancestry_with_ghosts(self, versions):
1580
"""See VersionedFile.get_ancestry."""
1581
if not self._parents:
1582
return self._parentless_ancestry(versions)
1583
# XXX: This will do len(history) index calls - perhaps
1584
# it should be altered to be a index core feature?
1585
# get a graph of all the mentioned versions:
1586
graph = {}
1587
versions = self._version_ids_to_keys(versions)
1588
pending = set(versions)
1589
while pending:
1590
# get all pending nodes
1591
this_iteration = pending
1592
new_nodes = self._get_entries(this_iteration)
1593
pending = set()
1594
for (index, key, value, node_refs) in new_nodes:
1595
graph[key] = node_refs[0]
1596
# queue parents
1597
for parent in graph[key]:
1598
# dont examine known nodes again
1599
if parent in graph:
1600
continue
1601
pending.add(parent)
1602
missing_versions = this_iteration.difference(graph)
1603
missing_needed = versions.intersection(missing_versions)
1604
if missing_needed:
1605
raise RevisionNotPresent(missing_needed.pop(), self)
1606
for missing_version in missing_versions:
1607
# add a key, no parents
1608
graph[missing_version] = []
1609
pending.discard(missing_version) # don't look for it
1610
result_keys = topo_sort(graph.items())
1611
return [key[0] for key in result_keys]
1612
1613
def get_graph(self):
1614
"""Return a list of the node:parents lists from this knit index."""
1615
if not self._parents:
1616
return [(key, ()) for key in self.get_versions()]
1617
result = []
1618
for index, key, value, refs in self._graph_index.iter_all_entries():
1619
result.append((key[0], tuple([ref[0] for ref in refs[0]])))
1620
return result
1621
1622
def iter_parents(self, version_ids):
1623
"""Iterate through the parents for many version ids.
1624
1625
:param version_ids: An iterable yielding version_ids.
1626
:return: An iterator that yields (version_id, parents). Requested
1627
version_ids not present in the versioned file are simply skipped.
1628
The order is undefined, allowing for different optimisations in
1629
the underlying implementation.
1630
"""
1631
if self._parents:
1632
all_nodes = set(self._get_entries(self._version_ids_to_keys(version_ids)))
1633
all_parents = set()
1634
present_parents = set()
1635
for node in all_nodes:
1636
all_parents.update(node[3][0])
1637
# any node we are querying must be present
1638
present_parents.add(node[1])
1639
unknown_parents = all_parents.difference(present_parents)
1640
present_parents.update(self._present_keys(unknown_parents))
1641
for node in all_nodes:
1642
parents = []
1643
for parent in node[3][0]:
1644
if parent in present_parents:
1645
parents.append(parent[0])
1646
yield node[1][0], tuple(parents)
1647
else:
1648
for node in self._get_entries(self._version_ids_to_keys(version_ids)):
1649
yield node[1][0], ()
1650
1651
def num_versions(self):
1652
return len(list(self._graph_index.iter_all_entries()))
1653
1654
__len__ = num_versions
1655
1656
def get_versions(self):
1657
"""Get all the versions in the file. not topologically sorted."""
1658
return [node[1][0] for node in self._graph_index.iter_all_entries()]
1659
1660
def has_version(self, version_id):
1661
"""True if the version is in the index."""
1662
return len(self._present_keys(self._version_ids_to_keys([version_id]))) == 1
1663
1664
def _keys_to_version_ids(self, keys):
1665
return tuple(key[0] for key in keys)
1666
1667
def get_position(self, version_id):
1668
"""Return details needed to access the version.
1669
1670
:return: a tuple (index, data position, size) to hand to the access
1671
logic to get the record.
1672
"""
1673
node = self._get_node(version_id)
1674
bits = node[2][1:].split(' ')
1675
return node[0], int(bits[0]), int(bits[1])
1676
1677
def get_method(self, version_id):
1678
"""Return compression method of specified version."""
1679
if not self._deltas:
1680
return 'fulltext'
1681
return self._parent_compression(self._get_node(version_id)[3][1])
1682
1683
def _parent_compression(self, reference_list):
1684
# use the second reference list to decide if this is delta'd or not.
1685
if len(reference_list):
1686
return 'line-delta'
1687
else:
1688
return 'fulltext'
1689
1690
def _get_node(self, version_id):
1691
try:
1692
return list(self._get_entries(self._version_ids_to_keys([version_id])))[0]
1693
except IndexError:
1694
raise RevisionNotPresent(version_id, self)
1695
1696
def get_options(self, version_id):
1697
"""Return a string represention options.
1698
1699
e.g. foo,bar
1700
"""
1701
node = self._get_node(version_id)
1702
if not self._deltas:
1703
options = ['fulltext']
1704
else:
1705
options = [self._parent_compression(node[3][1])]
1706
if node[2][0] == 'N':
1707
options.append('no-eol')
1708
return options
1709
1710
def get_parents(self, version_id):
1711
"""Return parents of specified version ignoring ghosts."""
1712
parents = list(self.iter_parents([version_id]))
1713
if not parents:
1714
# missing key
1715
raise errors.RevisionNotPresent(version_id, self)
1716
return parents[0][1]
1717
1718
def get_parents_with_ghosts(self, version_id):
1719
"""Return parents of specified version with ghosts."""
1720
nodes = list(self._get_entries(self._version_ids_to_keys([version_id]),
1721
check_present=True))
1722
if not self._parents:
1723
return ()
1724
return self._keys_to_version_ids(nodes[0][3][0])
1725
1726
def check_versions_present(self, version_ids):
1727
"""Check that all specified versions are present."""
1728
keys = self._version_ids_to_keys(version_ids)
1729
present = self._present_keys(keys)
1730
missing = keys.difference(present)
1731
if missing:
1732
raise RevisionNotPresent(missing.pop(), self)
1733
1734
def add_version(self, version_id, options, access_memo, parents):
1735
"""Add a version record to the index."""
1736
return self.add_versions(((version_id, options, access_memo, parents),))
1737
1738
def add_versions(self, versions, random_id=False):
1739
"""Add multiple versions to the index.
2053
1740
2054
1741
This function does not insert data into the Immutable GraphIndex
2055
1742
backing the KnitGraphIndex, instead it prepares data for insertion by
2056
1743
the caller and checks that it is safe to insert then calls
2057
1744
self._add_callback with the prepared GraphIndex nodes.
2058
1745
2059
:param records: a list of tuples:
2060
(key, options, access_memo, parents).
1746
:param versions: a list of tuples:
1747
(version_id, options, pos, size, parents).
2061
1748
:param random_id: If True the ids being added were randomly generated
2062
1749
and no check for existence will be performed.
2063
1750
"""
2065
1752
raise errors.ReadOnlyError(self)
2066
1753
# we hope there are no repositories with inconsistent parentage
2067
1754
# anymore.
1755
# check for dups
2068
1756
2069
1757
keys = {}
2070
for (key, options, access_memo, parents) in records:
2071
if self._parents:
2072
parents = tuple(parents)
1758
for (version_id, options, access_memo, parents) in versions:
2073
1759
index, pos, size = access_memo
1760
key = (version_id, )
1761
parents = tuple((parent, ) for parent in parents)
2074
1762
if 'no-eol' in options:
2075
1763
value = 'N'
2076
1764
else:
2093
1781
"in parentless index.")
2094
1782
node_refs = ()
2095
1783
keys[key] = (value, node_refs)
2096
# check for dups
2097
1784
if not random_id:
2098
1785
present_nodes = self._get_entries(keys)
2099
1786
for (index, key, value, node_refs) in present_nodes:
2100
if (value[0] != keys[key][0][0] or
2101
node_refs != keys[key][1]):
2102
raise KnitCorrupt(self, "inconsistent details in add_records"
1787
if (value, node_refs) != keys[key]:
1788
raise KnitCorrupt(self, "inconsistent details in add_versions"
2103
1789
": %s %s" % ((value, node_refs), keys[key]))
2104
1790
del keys[key]
2105
1791
result = []
2111
1797
result.append((key, value))
2112
1798
self._add_callback(result)
2113
1799
2114
def _check_read(self):
2115
"""raise if reads are not permitted."""
2116
if not self._is_locked():
2117
raise errors.ObjectNotLocked(self)
2118
2119
def _check_write_ok(self):
2120
"""Assert if writes are not permitted."""
2121
if not self._is_locked():
2122
raise errors.ObjectNotLocked(self)
2123
2124
def _compression_parent(self, an_entry):
2125
# return the key that an_entry is compressed against, or None
2126
# Grab the second parent list (as deltas implies parents currently)
2127
compression_parents = an_entry[3][1]
2128
if not compression_parents:
2129
return None
2130
if len(compression_parents) != 1:
2131
raise AssertionError(
2132
"Too many compression parents: %r" % compression_parents)
2133
return compression_parents[0]
2134
2135
def get_build_details(self, keys):
2136
"""Get the method, index_memo and compression parent for version_ids.
2137
2138
Ghosts are omitted from the result.
2139
2140
:param keys: An iterable of keys.
2141
:return: A dict of key:
2142
(index_memo, compression_parent, parents, record_details).
2143
index_memo
2144
opaque structure to pass to read_records to extract the raw
2145
data
2146
compression_parent
2147
Content that this record is built upon, may be None
2148
parents
2149
Logical parents of this node
2150
record_details
2151
extra information about the content which needs to be passed to
2152
Factory.parse_record
2153
"""
2154
self._check_read()
2155
result = {}
2156
entries = self._get_entries(keys, False)
2157
for entry in entries:
2158
key = entry[1]
2159
if not self._parents:
2160
parents = ()
2161
else:
2162
parents = entry[3][0]
2163
if not self._deltas:
2164
compression_parent_key = None
2165
else:
2166
compression_parent_key = self._compression_parent(entry)
2167
noeol = (entry[2][0] == 'N')
2168
if compression_parent_key:
2169
method = 'line-delta'
2170
else:
2171
method = 'fulltext'
2172
result[key] = (self._node_to_position(entry),
2173
compression_parent_key, parents,
2174
(method, noeol))
2175
return result
2176
2177
def _get_entries(self, keys, check_present=False):
2178
"""Get the entries for keys.
2179
2180
:param keys: An iterable of index key tuples.
2181
"""
2182
keys = set(keys)
2183
found_keys = set()
2184
if self._parents:
2185
for node in self._graph_index.iter_entries(keys):
2186
yield node
2187
found_keys.add(node[1])
2188
else:
2189
# adapt parentless index to the rest of the code.
2190
for node in self._graph_index.iter_entries(keys):
2191
yield node[0], node[1], node[2], ()
2192
found_keys.add(node[1])
2193
if check_present:
2194
missing_keys = keys.difference(found_keys)
2195
if missing_keys:
2196
raise RevisionNotPresent(missing_keys.pop(), self)
2197
2198
def get_method(self, key):
2199
"""Return compression method of specified key."""
2200
return self._get_method(self._get_node(key))
2201
2202
def _get_method(self, node):
2203
if not self._deltas:
2204
return 'fulltext'
2205
if self._compression_parent(node):
2206
return 'line-delta'
2207
else:
2208
return 'fulltext'
2209
2210
def _get_node(self, key):
2211
try:
2212
return list(self._get_entries([key]))[0]
2213
except IndexError:
2214
raise RevisionNotPresent(key, self)
2215
2216
def get_options(self, key):
2217
"""Return a list representing options.
2218
2219
e.g. ['foo', 'bar']
2220
"""
2221
node = self._get_node(key)
2222
options = [self._get_method(node)]
2223
if node[2][0] == 'N':
2224
options.append('no-eol')
2225
return options
2226
2227
def get_parent_map(self, keys):
2228
"""Get a map of the parents of keys.
2229
2230
:param keys: The keys to look up parents for.
2231
:return: A mapping from keys to parents. Absent keys are absent from
2232
the mapping.
2233
"""
2234
self._check_read()
2235
nodes = self._get_entries(keys)
2236
result = {}
2237
if self._parents:
2238
for node in nodes:
2239
result[node[1]] = node[3][0]
2240
else:
2241
for node in nodes:
2242
result[node[1]] = None
2243
return result
2244
2245
def get_position(self, key):
2246
"""Return details needed to access the version.
2247
2248
:return: a tuple (index, data position, size) to hand to the access
2249
logic to get the record.
2250
"""
2251
node = self._get_node(key)
2252
return self._node_to_position(node)
2253
2254
def keys(self):
2255
"""Get all the keys in the collection.
2256
2257
The keys are not ordered.
2258
"""
2259
self._check_read()
2260
return [node[1] for node in self._graph_index.iter_all_entries()]
2261
2262
def _node_to_position(self, node):
2263
"""Convert an index value to position details."""
2264
bits = node[2][1:].split(' ')
2265
return node[0], int(bits[0]), int(bits[1])
2266
2267
2268
class _KnitKeyAccess(object):
2269
"""Access to records in .knit files."""
2270
2271
def __init__(self, transport, mapper):
2272
"""Create a _KnitKeyAccess with transport and mapper.
2273
2274
:param transport: The transport the access object is rooted at.
2275
:param mapper: The mapper used to map keys to .knit files.
1800
def _version_ids_to_keys(self, version_ids):
1801
return set((version_id, ) for version_id in version_ids)
1802
1803
1804
class _KnitAccess(object):
1805
"""Access to knit records in a .knit file."""
1806
1807
def __init__(self, transport, filename, _file_mode, _dir_mode,
1808
_need_to_create, _create_parent_dir):
1809
"""Create a _KnitAccess for accessing and inserting data.
1810
1811
:param transport: The transport the .knit is located on.
1812
:param filename: The filename of the .knit.
2276
1813
"""
2277
1814
self._transport = transport
2278
self._mapper = mapper
1815
self._filename = filename
1816
self._file_mode = _file_mode
1817
self._dir_mode = _dir_mode
1818
self._need_to_create = _need_to_create
1819
self._create_parent_dir = _create_parent_dir
2279
1820
2280
def add_raw_records(self, key_sizes, raw_data):
1821
def add_raw_records(self, sizes, raw_data):
2281
1822
"""Add raw knit bytes to a storage area.
2282
1823
2283
The data is spooled to the container writer in one bytes-record per
2284
raw data item.
1824
The data is spooled to whereever the access method is storing data.
2285
1825
2286
:param sizes: An iterable of tuples containing the key and size of each
2287
raw data segment.
1826
:param sizes: An iterable containing the size of each raw data segment.
2288
1827
:param raw_data: A bytestring containing the data.
2289
:return: A list of memos to retrieve the record later. Each memo is an
2290
opaque index memo. For _KnitKeyAccess the memo is (key, pos,
2291
length), where the key is the record key.
1828
:return: A list of memos to retrieve the record later. Each memo is a
1829
tuple - (index, pos, length), where the index field is always None
1830
for the .knit access method.
2292
1831
"""
2293
if type(raw_data) != str:
2294
raise AssertionError(
2295
'data must be plain bytes was %s' % type(raw_data))
1832
assert type(raw_data) == str, \
1833
'data must be plain bytes was %s' % type(raw_data)
1834
if not self._need_to_create:
1835
base = self._transport.append_bytes(self._filename, raw_data)
1836
else:
1837
self._transport.put_bytes_non_atomic(self._filename, raw_data,
1838
create_parent_dir=self._create_parent_dir,
1839
mode=self._file_mode,
1840
dir_mode=self._dir_mode)
1841
self._need_to_create = False
1842
base = 0
2296
1843
result = []
2297
offset = 0
2298
# TODO: This can be tuned for writing to sftp and other servers where
2299
# append() is relatively expensive by grouping the writes to each key
2300
# prefix.
2301
for key, size in key_sizes:
2302
path = self._mapper.map(key)
2303
try:
2304
base = self._transport.append_bytes(path + '.knit',
2305
raw_data[offset:offset+size])
2306
except errors.NoSuchFile:
2307
self._transport.mkdir(osutils.dirname(path))
2308
base = self._transport.append_bytes(path + '.knit',
2309
raw_data[offset:offset+size])
2310
# if base == 0:
2311
# chmod.
2312
offset += size
2313
result.append((key, base, size))
1844
for size in sizes:
1845
result.append((None, base, size))
1846
base += size
2314
1847
return result
2315
1848
1849
def create(self):
1850
"""IFF this data access has its own storage area, initialise it.
1851
1852
:return: None.
1853
"""
1854
self._transport.put_bytes_non_atomic(self._filename, '',
1855
mode=self._file_mode)
1856
1857
def open_file(self):
1858
"""IFF this data access can be represented as a single file, open it.
1859
1860
For knits that are not mapped to a single file on disk this will
1861
always return None.
1862
1863
:return: None or a file handle.
1864
"""
1865
try:
1866
return self._transport.get(self._filename)
1867
except NoSuchFile:
1868
pass
1869
return None
1870
2316
1871
def get_raw_records(self, memos_for_retrieval):
2317
1872
"""Get the raw bytes for a records.
2318
1873
2319
:param memos_for_retrieval: An iterable containing the access memo for
2320
retrieving the bytes.
1874
:param memos_for_retrieval: An iterable containing the (index, pos,
1875
length) memo for retrieving the bytes. The .knit method ignores
1876
the index as there is always only a single file.
2321
1877
:return: An iterator over the bytes of the records.
2322
1878
"""
2323
# first pass, group into same-index request to minimise readv's issued.
2324
request_lists = []
2325
current_prefix = None
2326
for (key, offset, length) in memos_for_retrieval:
2327
if current_prefix == key[:-1]:
2328
current_list.append((offset, length))
2329
else:
2330
if current_prefix is not None:
2331
request_lists.append((current_prefix, current_list))
2332
current_prefix = key[:-1]
2333
current_list = [(offset, length)]
2334
# handle the last entry
2335
if current_prefix is not None:
2336
request_lists.append((current_prefix, current_list))
2337
for prefix, read_vector in request_lists:
2338
path = self._mapper.map(prefix) + '.knit'
2339
for pos, data in self._transport.readv(path, read_vector):
2340
yield data
2341
2342
2343
class _DirectPackAccess(object):
2344
"""Access to data in one or more packs with less translation."""
2345
2346
def __init__(self, index_to_packs):
2347
"""Create a _DirectPackAccess object.
1879
read_vector = [(pos, size) for (index, pos, size) in memos_for_retrieval]
1880
for pos, data in self._transport.readv(self._filename, read_vector):
1881
yield data
1882
1883
1884
class _PackAccess(object):
1885
"""Access to knit records via a collection of packs."""
1886
1887
def __init__(self, index_to_packs, writer=None):
1888
"""Create a _PackAccess object.
2348
1889
2349
1890
:param index_to_packs: A dict mapping index objects to the transport
2350
1891
and file names for obtaining data.
1892
:param writer: A tuple (pack.ContainerWriter, write_index) which
1893
contains the pack to write, and the index that reads from it will
1894
be associated with.
2351
1895
"""
2352
self._container_writer = None
2353
self._write_index = None
2354
self._indices = index_to_packs
1896
if writer:
1897
self.container_writer = writer[0]
1898
self.write_index = writer[1]
1899
else:
1900
self.container_writer = None
1901
self.write_index = None
1902
self.indices = index_to_packs
2355
1903
2356
def add_raw_records(self, key_sizes, raw_data):
1904
def add_raw_records(self, sizes, raw_data):
2357
1905
"""Add raw knit bytes to a storage area.
2358
1906
2359
1907
The data is spooled to the container writer in one bytes-record per
2360
1908
raw data item.
2361
1909
2362
:param sizes: An iterable of tuples containing the key and size of each
2363
raw data segment.
1910
:param sizes: An iterable containing the size of each raw data segment.
2364
1911
:param raw_data: A bytestring containing the data.
2365
:return: A list of memos to retrieve the record later. Each memo is an
2366
opaque index memo. For _DirectPackAccess the memo is (index, pos,
2367
length), where the index field is the write_index object supplied
2368
to the PackAccess object.
1912
:return: A list of memos to retrieve the record later. Each memo is a
1913
tuple - (index, pos, length), where the index field is the
1914
write_index object supplied to the PackAccess object.
2369
1915
"""
2370
if type(raw_data) != str:
2371
raise AssertionError(
2372
'data must be plain bytes was %s' % type(raw_data))
1916
assert type(raw_data) == str, \
1917
'data must be plain bytes was %s' % type(raw_data)
2373
1918
result = []
2374
1919
offset = 0
2375
for key, size in key_sizes:
2376
p_offset, p_length = self._container_writer.add_bytes_record(
1920
for size in sizes:
1921
p_offset, p_length = self.container_writer.add_bytes_record(
2377
1922
raw_data[offset:offset+size], [])
2378
1923
offset += size
2379
result.append((self._write_index, p_offset, p_length))
1924
result.append((self.write_index, p_offset, p_length))
2380
1925
return result
2381
1926
1927
def create(self):
1928
"""Pack based knits do not get individually created."""
1929
2382
1930
def get_raw_records(self, memos_for_retrieval):
2383
1931
"""Get the raw bytes for a records.
2384
1932
2403
1951
if current_index is not None:
2404
1952
request_lists.append((current_index, current_list))
2405
1953
for index, offsets in request_lists:
2406
transport, path = self._indices[index]
1954
transport, path = self.indices[index]
2407
1955
reader = pack.make_readv_reader(transport, path, offsets)
2408
1956
for names, read_func in reader.iter_records():
2409
1957
yield read_func(None)
2410
1958
2411
def set_writer(self, writer, index, transport_packname):
1959
def open_file(self):
1960
"""Pack based knits have no single file."""
1961
return None
1962
1963
def set_writer(self, writer, index, (transport, packname)):
2412
1964
"""Set a writer to use for adding data."""
2413
if index is not None:
2414
self._indices[index] = transport_packname
2415
self._container_writer = writer
2416
self._write_index = index
1965
self.indices[index] = (transport, packname)
1966
self.container_writer = writer
1967
self.write_index = index
1968
1969
1970
class _KnitData(object):
1971
"""Manage extraction of data from a KnitAccess, caching and decompressing.
1972
1973
The KnitData class provides the logic for parsing and using knit records,
1974
making use of an access method for the low level read and write operations.
1975
"""
1976
1977
def __init__(self, access):
1978
"""Create a KnitData object.
1979
1980
:param access: The access method to use. Access methods such as
1981
_KnitAccess manage the insertion of raw records and the subsequent
1982
retrieval of the same.
1983
"""
1984
self._access = access
1985
self._checked = False
1986
# TODO: jam 20060713 conceptually, this could spill to disk
1987
# if the cached size gets larger than a certain amount
1988
# but it complicates the model a bit, so for now just use
1989
# a simple dictionary
1990
self._cache = {}
1991
self._do_cache = False
1992
1993
def enable_cache(self):
1994
"""Enable caching of reads."""
1995
self._do_cache = True
1996
1997
def clear_cache(self):
1998
"""Clear the record cache."""
1999
self._do_cache = False
2000
self._cache = {}
2001
2002
def _open_file(self):
2003
return self._access.open_file()
2004
2005
def _record_to_data(self, version_id, digest, lines, dense_lines=None):
2006
"""Convert version_id, digest, lines into a raw data block.
2007
2008
:param dense_lines: The bytes of lines but in a denser form. For
2009
instance, if lines is a list of 1000 bytestrings each ending in \n,
2010
dense_lines may be a list with one line in it, containing all the
2011
1000's lines and their \n's. Using dense_lines if it is already
2012
known is a win because the string join to create bytes in this
2013
function spends less time resizing the final string.
2014
:return: (len, a StringIO instance with the raw data ready to read.)
2015
"""
2016
# Note: using a string copy here increases memory pressure with e.g.
2017
# ISO's, but it is about 3 seconds faster on a 1.2Ghz intel machine
2018
# when doing the initial commit of a mozilla tree. RBC 20070921
2019
bytes = ''.join(chain(
2020
["version %s %d %s\n" % (version_id,
2021
len(lines),
2022
digest)],
2023
dense_lines or lines,
2024
["end %s\n" % version_id]))
2025
assert bytes.__class__ == str
2026
compressed_bytes = bytes_to_gzip(bytes)
2027
return len(compressed_bytes), compressed_bytes
2028
2029
def add_raw_records(self, sizes, raw_data):
2030
"""Append a prepared record to the data file.
2031
2032
:param sizes: An iterable containing the size of each raw data segment.
2033
:param raw_data: A bytestring containing the data.
2034
:return: a list of index data for the way the data was stored.
2035
See the access method add_raw_records documentation for more
2036
details.
2037
"""
2038
return self._access.add_raw_records(sizes, raw_data)
2039
2040
def _parse_record_header(self, version_id, raw_data):
2041
"""Parse a record header for consistency.
2042
2043
:return: the header and the decompressor stream.
2044
as (stream, header_record)
2045
"""
2046
df = GzipFile(mode='rb', fileobj=StringIO(raw_data))
2047
try:
2048
rec = self._check_header(version_id, df.readline())
2049
except Exception, e:
2050
raise KnitCorrupt(self._access,
2051
"While reading {%s} got %s(%s)"
2052
% (version_id, e.__class__.__name__, str(e)))
2053
return df, rec
2054
2055
def _check_header(self, version_id, line):
2056
rec = line.split()
2057
if len(rec) != 4:
2058
raise KnitCorrupt(self._access,
2059
'unexpected number of elements in record header')
2060
if rec[1] != version_id:
2061
raise KnitCorrupt(self._access,
2062
'unexpected version, wanted %r, got %r'
2063
% (version_id, rec[1]))
2064
return rec
2065
2066
def _parse_record(self, version_id, data):
2067
# profiling notes:
2068
# 4168 calls in 2880 217 internal
2069
# 4168 calls to _parse_record_header in 2121
2070
# 4168 calls to readlines in 330
2071
df = GzipFile(mode='rb', fileobj=StringIO(data))
2072
2073
try:
2074
record_contents = df.readlines()
2075
except Exception, e:
2076
raise KnitCorrupt(self._access,
2077
"While reading {%s} got %s(%s)"
2078
% (version_id, e.__class__.__name__, str(e)))
2079
header = record_contents.pop(0)
2080
rec = self._check_header(version_id, header)
2081
2082
last_line = record_contents.pop()
2083
if len(record_contents) != int(rec[2]):
2084
raise KnitCorrupt(self._access,
2085
'incorrect number of lines %s != %s'
2086
' for version {%s}'
2087
% (len(record_contents), int(rec[2]),
2088
version_id))
2089
if last_line != 'end %s\n' % rec[1]:
2090
raise KnitCorrupt(self._access,
2091
'unexpected version end line %r, wanted %r'
2092
% (last_line, version_id))
2093
df.close()
2094
return record_contents, rec[3]
2095
2096
def read_records_iter_raw(self, records):
2097
"""Read text records from data file and yield raw data.
2098
2099
This unpacks enough of the text record to validate the id is
2100
as expected but thats all.
2101
"""
2102
# setup an iterator of the external records:
2103
# uses readv so nice and fast we hope.
2104
if len(records):
2105
# grab the disk data needed.
2106
if self._cache:
2107
# Don't check _cache if it is empty
2108
needed_offsets = [index_memo for version_id, index_memo
2109
in records
2110
if version_id not in self._cache]
2111
else:
2112
needed_offsets = [index_memo for version_id, index_memo
2113
in records]
2114
2115
raw_records = self._access.get_raw_records(needed_offsets)
2116
2117
for version_id, index_memo in records:
2118
if version_id in self._cache:
2119
# This data has already been validated
2120
data = self._cache[version_id]
2121
else:
2122
data = raw_records.next()
2123
if self._do_cache:
2124
self._cache[version_id] = data
2125
2126
# validate the header
2127
df, rec = self._parse_record_header(version_id, data)
2128
df.close()
2129
yield version_id, data
2130
2131
def read_records_iter(self, records):
2132
"""Read text records from data file and yield result.
2133
2134
The result will be returned in whatever is the fastest to read.
2135
Not by the order requested. Also, multiple requests for the same
2136
record will only yield 1 response.
2137
:param records: A list of (version_id, pos, len) entries
2138
:return: Yields (version_id, contents, digest) in the order
2139
read, not the order requested
2140
"""
2141
if not records:
2142
return
2143
2144
if self._cache:
2145
# Skip records we have alread seen
2146
yielded_records = set()
2147
needed_records = set()
2148
for record in records:
2149
if record[0] in self._cache:
2150
if record[0] in yielded_records:
2151
continue
2152
yielded_records.add(record[0])
2153
data = self._cache[record[0]]
2154
content, digest = self._parse_record(record[0], data)
2155
yield (record[0], content, digest)
2156
else:
2157
needed_records.add(record)
2158
needed_records = sorted(needed_records, key=operator.itemgetter(1))
2159
else:
2160
needed_records = sorted(set(records), key=operator.itemgetter(1))
2161
2162
if not needed_records:
2163
return
2164
2165
# The transport optimizes the fetching as well
2166
# (ie, reads continuous ranges.)
2167
raw_data = self._access.get_raw_records(
2168
[index_memo for version_id, index_memo in needed_records])
2169
2170
for (version_id, index_memo), data in \
2171
izip(iter(needed_records), raw_data):
2172
content, digest = self._parse_record(version_id, data)
2173
if self._do_cache:
2174
self._cache[version_id] = data
2175
yield version_id, content, digest
2176
2177
def read_records(self, records):
2178
"""Read records into a dictionary."""
2179
components = {}
2180
for record_id, content, digest in \
2181
self.read_records_iter(records):
2182
components[record_id] = (content, digest)
2183
return components
2184
2185
2186
class InterKnit(InterVersionedFile):
2187
"""Optimised code paths for knit to knit operations."""
2188
2189
_matching_file_from_factory = KnitVersionedFile
2190
_matching_file_to_factory = KnitVersionedFile
2191
2192
@staticmethod
2193
def is_compatible(source, target):
2194
"""Be compatible with knits. """
2195
try:
2196
return (isinstance(source, KnitVersionedFile) and
2197
isinstance(target, KnitVersionedFile))
2198
except AttributeError:
2199
return False
2200
2201
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2202
"""See InterVersionedFile.join."""
2203
assert isinstance(self.source, KnitVersionedFile)
2204
assert isinstance(self.target, KnitVersionedFile)
2205
2206
# If the source and target are mismatched w.r.t. annotations vs
2207
# plain, the data needs to be converted accordingly
2208
if self.source.factory.annotated == self.target.factory.annotated:
2209
converter = None
2210
elif self.source.factory.annotated:
2211
converter = self._anno_to_plain_converter
2212
else:
2213
# We're converting from a plain to an annotated knit. This requires
2214
# building the annotations from scratch. The generic join code
2215
# handles this implicitly so we delegate to it.
2216
return super(InterKnit, self).join(pb, msg, version_ids,
2217
ignore_missing)
2218
2219
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2220
if not version_ids:
2221
return 0
2222
2223
pb = ui.ui_factory.nested_progress_bar()
2224
try:
2225
version_ids = list(version_ids)
2226
if None in version_ids:
2227
version_ids.remove(None)
2228
2229
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2230
this_versions = set(self.target._index.get_versions())
2231
# XXX: For efficiency we should not look at the whole index,
2232
# we only need to consider the referenced revisions - they
2233
# must all be present, or the method must be full-text.
2234
# TODO, RBC 20070919
2235
needed_versions = self.source_ancestry - this_versions
2236
2237
if not needed_versions:
2238
return 0
2239
full_list = topo_sort(self.source.get_graph())
2240
2241
version_list = [i for i in full_list if (not self.target.has_version(i)
2242
and i in needed_versions)]
2243
2244
# plan the join:
2245
copy_queue = []
2246
copy_queue_records = []
2247
copy_set = set()
2248
for version_id in version_list:
2249
options = self.source._index.get_options(version_id)
2250
parents = self.source._index.get_parents_with_ghosts(version_id)
2251
# check that its will be a consistent copy:
2252
for parent in parents:
2253
# if source has the parent, we must :
2254
# * already have it or
2255
# * have it scheduled already
2256
# otherwise we don't care
2257
assert (self.target.has_version(parent) or
2258
parent in copy_set or
2259
not self.source.has_version(parent))
2260
index_memo = self.source._index.get_position(version_id)
2261
copy_queue_records.append((version_id, index_memo))
2262
copy_queue.append((version_id, options, parents))
2263
copy_set.add(version_id)
2264
2265
# data suck the join:
2266
count = 0
2267
total = len(version_list)
2268
raw_datum = []
2269
raw_records = []
2270
for (version_id, raw_data), \
2271
(version_id2, options, parents) in \
2272
izip(self.source._data.read_records_iter_raw(copy_queue_records),
2273
copy_queue):
2274
assert version_id == version_id2, 'logic error, inconsistent results'
2275
count = count + 1
2276
pb.update("Joining knit", count, total)
2277
if converter:
2278
size, raw_data = converter(raw_data, version_id, options,
2279
parents)
2280
else:
2281
size = len(raw_data)
2282
raw_records.append((version_id, options, parents, size))
2283
raw_datum.append(raw_data)
2284
self.target._add_raw_records(raw_records, ''.join(raw_datum))
2285
return count
2286
finally:
2287
pb.finished()
2288
2289
def _anno_to_plain_converter(self, raw_data, version_id, options,
2290
parents):
2291
"""Convert annotated content to plain content."""
2292
data, digest = self.source._data._parse_record(version_id, raw_data)
2293
if 'fulltext' in options:
2294
content = self.source.factory.parse_fulltext(data, version_id)
2295
lines = self.target.factory.lower_fulltext(content)
2296
else:
2297
delta = self.source.factory.parse_line_delta(data, version_id,
2298
plain=True)
2299
lines = self.target.factory.lower_line_delta(delta)
2300
return self.target._data._record_to_data(version_id, digest, lines)
2301
2302
2303
InterVersionedFile.register_optimiser(InterKnit)
2304
2305
2306
class WeaveToKnit(InterVersionedFile):
2307
"""Optimised code paths for weave to knit operations."""
2308
2309
_matching_file_from_factory = bzrlib.weave.WeaveFile
2310
_matching_file_to_factory = KnitVersionedFile
2311
2312
@staticmethod
2313
def is_compatible(source, target):
2314
"""Be compatible with weaves to knits."""
2315
try:
2316
return (isinstance(source, bzrlib.weave.Weave) and
2317
isinstance(target, KnitVersionedFile))
2318
except AttributeError:
2319
return False
2320
2321
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
2322
"""See InterVersionedFile.join."""
2323
assert isinstance(self.source, bzrlib.weave.Weave)
2324
assert isinstance(self.target, KnitVersionedFile)
2325
2326
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
2327
2328
if not version_ids:
2329
return 0
2330
2331
pb = ui.ui_factory.nested_progress_bar()
2332
try:
2333
version_ids = list(version_ids)
2334
2335
self.source_ancestry = set(self.source.get_ancestry(version_ids))
2336
this_versions = set(self.target._index.get_versions())
2337
needed_versions = self.source_ancestry - this_versions
2338
2339
if not needed_versions:
2340
return 0
2341
full_list = topo_sort(self.source.get_graph())
2342
2343
version_list = [i for i in full_list if (not self.target.has_version(i)
2344
and i in needed_versions)]
2345
2346
# do the join:
2347
count = 0
2348
total = len(version_list)
2349
for version_id in version_list:
2350
pb.update("Converting to knit", count, total)
2351
parents = self.source.get_parents(version_id)
2352
# check that its will be a consistent copy:
2353
for parent in parents:
2354
# if source has the parent, we must already have it
2355
assert (self.target.has_version(parent))
2356
self.target.add_lines(
2357
version_id, parents, self.source.get_lines(version_id))
2358
count = count + 1
2359
return count
2360
finally:
2361
pb.finished()
2362
2363
2364
InterVersionedFile.register_optimiser(WeaveToKnit)
2417
2365
2418
2366
2419
2367
# Deprecated, use PatienceSequenceMatcher instead
2427
2375
It will work for knits with cached annotations, but this is not
2428
2376
recommended.
2429
2377
"""
2430
annotator = _KnitAnnotator(knit)
2431
return iter(annotator.annotate(revision_id))
2432
2433
2434
class _KnitAnnotator(object):
2435
"""Build up the annotations for a text."""
2436
2437
def __init__(self, knit):
2438
self._knit = knit
2439
2440
# Content objects, differs from fulltexts because of how final newlines
2441
# are treated by knits. the content objects here will always have a
2442
# final newline
2443
self._fulltext_contents = {}
2444
2445
# Annotated lines of specific revisions
2446
self._annotated_lines = {}
2447
2448
# Track the raw data for nodes that we could not process yet.
2449
# This maps the revision_id of the base to a list of children that will
2450
# annotated from it.
2451
self._pending_children = {}
2452
2453
# Nodes which cannot be extracted
2454
self._ghosts = set()
2455
2456
# Track how many children this node has, so we know if we need to keep
2457
# it
2458
self._annotate_children = {}
2459
self._compression_children = {}
2460
2461
self._all_build_details = {}
2462
# The children => parent revision_id graph
2463
self._revision_id_graph = {}
2464
2465
self._heads_provider = None
2466
2467
self._nodes_to_keep_annotations = set()
2468
self._generations_until_keep = 100
2469
2470
def set_generations_until_keep(self, value):
2471
"""Set the number of generations before caching a node.
2472
2473
Setting this to -1 will cache every merge node, setting this higher
2474
will cache fewer nodes.
2475
"""
2476
self._generations_until_keep = value
2477
2478
def _add_fulltext_content(self, revision_id, content_obj):
2479
self._fulltext_contents[revision_id] = content_obj
2480
# TODO: jam 20080305 It might be good to check the sha1digest here
2481
return content_obj.text()
2482
2483
def _check_parents(self, child, nodes_to_annotate):
2484
"""Check if all parents have been processed.
2485
2486
:param child: A tuple of (rev_id, parents, raw_content)
2487
:param nodes_to_annotate: If child is ready, add it to
2488
nodes_to_annotate, otherwise put it back in self._pending_children
2489
"""
2490
for parent_id in child[1]:
2491
if (parent_id not in self._annotated_lines):
2492
# This parent is present, but another parent is missing
2493
self._pending_children.setdefault(parent_id,
2494
[]).append(child)
2495
break
2378
ancestry = knit.get_ancestry(revision_id)
2379
fulltext = dict(zip(ancestry, knit.get_line_list(ancestry)))
2380
annotations = {}
2381
for candidate in ancestry:
2382
if candidate in annotations:
2383
continue
2384
parents = knit.get_parents(candidate)
2385
if len(parents) == 0:
2386
blocks = None
2387
elif knit._index.get_method(candidate) != 'line-delta':
2388
blocks = None
2496
2389
else:
2497
# This one is ready to be processed
2498
nodes_to_annotate.append(child)
2499
2500
def _add_annotation(self, revision_id, fulltext, parent_ids,
2501
left_matching_blocks=None):
2502
"""Add an annotation entry.
2503
2504
All parents should already have been annotated.
2505
:return: A list of children that now have their parents satisfied.
2506
"""
2507
a = self._annotated_lines
2508
annotated_parent_lines = [a[p] for p in parent_ids]
2509
annotated_lines = list(annotate.reannotate(annotated_parent_lines,
2510
fulltext, revision_id, left_matching_blocks,
2511
heads_provider=self._get_heads_provider()))
2512
self._annotated_lines[revision_id] = annotated_lines
2513
for p in parent_ids:
2514
ann_children = self._annotate_children[p]
2515
ann_children.remove(revision_id)
2516
if (not ann_children
2517
and p not in self._nodes_to_keep_annotations):
2518
del self._annotated_lines[p]
2519
del self._all_build_details[p]
2520
if p in self._fulltext_contents:
2521
del self._fulltext_contents[p]
2522
# Now that we've added this one, see if there are any pending
2523
# deltas to be done, certainly this parent is finished
2524
nodes_to_annotate = []
2525
for child in self._pending_children.pop(revision_id, []):
2526
self._check_parents(child, nodes_to_annotate)
2527
return nodes_to_annotate
2528
2529
def _get_build_graph(self, key):
2530
"""Get the graphs for building texts and annotations.
2531
2532
The data you need for creating a full text may be different than the
2533
data you need to annotate that text. (At a minimum, you need both
2534
parents to create an annotation, but only need 1 parent to generate the
2535
fulltext.)
2536
2537
:return: A list of (key, index_memo) records, suitable for
2538
passing to read_records_iter to start reading in the raw data fro/
2539
the pack file.
2540
"""
2541
if key in self._annotated_lines:
2542
# Nothing to do
2543
return []
2544
pending = set([key])
2545
records = []
2546
generation = 0
2547
kept_generation = 0
2548
while pending:
2549
# get all pending nodes
2550
generation += 1
2551
this_iteration = pending
2552
build_details = self._knit._index.get_build_details(this_iteration)
2553
self._all_build_details.update(build_details)
2554
# new_nodes = self._knit._index._get_entries(this_iteration)
2555
pending = set()
2556
for key, details in build_details.iteritems():
2557
(index_memo, compression_parent, parents,
2558
record_details) = details
2559
self._revision_id_graph[key] = parents
2560
records.append((key, index_memo))
2561
# Do we actually need to check _annotated_lines?
2562
pending.update(p for p in parents
2563
if p not in self._all_build_details)
2564
if compression_parent:
2565
self._compression_children.setdefault(compression_parent,
2566
[]).append(key)
2567
if parents:
2568
for parent in parents:
2569
self._annotate_children.setdefault(parent,
2570
[]).append(key)
2571
num_gens = generation - kept_generation
2572
if ((num_gens >= self._generations_until_keep)
2573
and len(parents) > 1):
2574
kept_generation = generation
2575
self._nodes_to_keep_annotations.add(key)
2576
2577
missing_versions = this_iteration.difference(build_details.keys())
2578
self._ghosts.update(missing_versions)
2579
for missing_version in missing_versions:
2580
# add a key, no parents
2581
self._revision_id_graph[missing_version] = ()
2582
pending.discard(missing_version) # don't look for it
2583
if self._ghosts.intersection(self._compression_children):
2584
raise KnitCorrupt(
2585
"We cannot have nodes which have a ghost compression parent:\n"
2586
"ghosts: %r\n"
2587
"compression children: %r"
2588
% (self._ghosts, self._compression_children))
2589
# Cleanout anything that depends on a ghost so that we don't wait for
2590
# the ghost to show up
2591
for node in self._ghosts:
2592
if node in self._annotate_children:
2593
# We won't be building this node
2594
del self._annotate_children[node]
2595
# Generally we will want to read the records in reverse order, because
2596
# we find the parent nodes after the children
2597
records.reverse()
2598
return records
2599
2600
def _annotate_records(self, records):
2601
"""Build the annotations for the listed records."""
2602
# We iterate in the order read, rather than a strict order requested
2603
# However, process what we can, and put off to the side things that
2604
# still need parents, cleaning them up when those parents are
2605
# processed.
2606
for (rev_id, record,
2607
digest) in self._knit._read_records_iter(records):
2608
if rev_id in self._annotated_lines:
2609
continue
2610
parent_ids = self._revision_id_graph[rev_id]
2611
parent_ids = [p for p in parent_ids if p not in self._ghosts]
2612
details = self._all_build_details[rev_id]
2613
(index_memo, compression_parent, parents,
2614
record_details) = details
2615
nodes_to_annotate = []
2616
# TODO: Remove the punning between compression parents, and
2617
# parent_ids, we should be able to do this without assuming
2618
# the build order
2619
if len(parent_ids) == 0:
2620
# There are no parents for this node, so just add it
2621
# TODO: This probably needs to be decoupled
2622
fulltext_content, delta = self._knit._factory.parse_record(
2623
rev_id, record, record_details, None)
2624
fulltext = self._add_fulltext_content(rev_id, fulltext_content)
2625
nodes_to_annotate.extend(self._add_annotation(rev_id, fulltext,
2626
parent_ids, left_matching_blocks=None))
2627
else:
2628
child = (rev_id, parent_ids, record)
2629
# Check if all the parents are present
2630
self._check_parents(child, nodes_to_annotate)
2631
while nodes_to_annotate:
2632
# Should we use a queue here instead of a stack?
2633
(rev_id, parent_ids, record) = nodes_to_annotate.pop()
2634
(index_memo, compression_parent, parents,
2635
record_details) = self._all_build_details[rev_id]
2636
if compression_parent is not None:
2637
comp_children = self._compression_children[compression_parent]
2638
if rev_id not in comp_children:
2639
raise AssertionError("%r not in compression children %r"
2640
% (rev_id, comp_children))
2641
# If there is only 1 child, it is safe to reuse this
2642
# content
2643
reuse_content = (len(comp_children) == 1
2644
and compression_parent not in
2645
self._nodes_to_keep_annotations)
2646
if reuse_content:
2647
# Remove it from the cache since it will be changing
2648
parent_fulltext_content = self._fulltext_contents.pop(compression_parent)
2649
# Make sure to copy the fulltext since it might be
2650
# modified
2651
parent_fulltext = list(parent_fulltext_content.text())
2652
else:
2653
parent_fulltext_content = self._fulltext_contents[compression_parent]
2654
parent_fulltext = parent_fulltext_content.text()
2655
comp_children.remove(rev_id)
2656
fulltext_content, delta = self._knit._factory.parse_record(
2657
rev_id, record, record_details,
2658
parent_fulltext_content,
2659
copy_base_content=(not reuse_content))
2660
fulltext = self._add_fulltext_content(rev_id,
2661
fulltext_content)
2662
blocks = KnitContent.get_line_delta_blocks(delta,
2663
parent_fulltext, fulltext)
2664
else:
2665
fulltext_content = self._knit._factory.parse_fulltext(
2666
record, rev_id)
2667
fulltext = self._add_fulltext_content(rev_id,
2668
fulltext_content)
2669
blocks = None
2670
nodes_to_annotate.extend(
2671
self._add_annotation(rev_id, fulltext, parent_ids,
2672
left_matching_blocks=blocks))
2673
2674
def _get_heads_provider(self):
2675
"""Create a heads provider for resolving ancestry issues."""
2676
if self._heads_provider is not None:
2677
return self._heads_provider
2678
parent_provider = _mod_graph.DictParentsProvider(
2679
self._revision_id_graph)
2680
graph_obj = _mod_graph.Graph(parent_provider)
2681
head_cache = _mod_graph.FrozenHeadsCache(graph_obj)
2682
self._heads_provider = head_cache
2683
return head_cache
2684
2685
def annotate(self, key):
2686
"""Return the annotated fulltext at the given key.
2687
2688
:param key: The key to annotate.
2689
"""
2690
if True or len(self._knit._fallback_vfs) > 0:
2691
# stacked knits can't use the fast path at present.
2692
return self._simple_annotate(key)
2693
records = self._get_build_graph(key)
2694
if key in self._ghosts:
2695
raise errors.RevisionNotPresent(key, self._knit)
2696
self._annotate_records(records)
2697
return self._annotated_lines[key]
2698
2699
def _simple_annotate(self, key):
2700
"""Return annotated fulltext, rediffing from the full texts.
2701
2702
This is slow but makes no assumptions about the repository
2703
being able to produce line deltas.
2704
"""
2705
# TODO: this code generates a parent maps of present ancestors; it
2706
# could be split out into a separate method, and probably should use
2707
# iter_ancestry instead. -- mbp and robertc 20080704
2708
graph = _mod_graph.Graph(self._knit)
2709
head_cache = _mod_graph.FrozenHeadsCache(graph)
2710
search = graph._make_breadth_first_searcher([key])
2711
keys = set()
2712
while True:
2713
try:
2714
present, ghosts = search.next_with_ghosts()
2715
except StopIteration:
2716
break
2717
keys.update(present)
2718
parent_map = self._knit.get_parent_map(keys)
2719
parent_cache = {}
2720
reannotate = annotate.reannotate
2721
for record in self._knit.get_record_stream(keys, 'topological', True):
2722
key = record.key
2723
fulltext = split_lines(record.get_bytes_as('fulltext'))
2724
parents = parent_map[key]
2725
if parents is not None:
2726
parent_lines = [parent_cache[parent] for parent in parent_map[key]]
2727
else:
2728
parent_lines = []
2729
parent_cache[key] = list(
2730
reannotate(parent_lines, fulltext, key, None, head_cache))
2731
try:
2732
return parent_cache[key]
2733
except KeyError, e:
2734
raise errors.RevisionNotPresent(key, self._knit)
2390
parent, sha1, noeol, delta = knit.get_delta(candidate)
2391
blocks = KnitContent.get_line_delta_blocks(delta,
2392
fulltext[parents[0]], fulltext[candidate])
2393
annotations[candidate] = list(annotate.reannotate([annotations[p]
2394
for p in parents], fulltext[candidate], candidate, blocks))
2395
return iter(annotations[revision_id])
2735
2396
2736
2397
2737
2398
try:
Loggerhead is a web-based interface for Breezy
Version: 2.0.1