151
139
texts/deltas (via (fileid, revisionid) tuples).
152
140
:param signature_index: A GraphIndex for determining what signatures are
153
141
present in the Pack and accessing the locations of their texts.
154
:param chk_index: A GraphIndex for accessing content by CHK, if the
157
143
self.revision_index = revision_index
158
144
self.inventory_index = inventory_index
159
145
self.text_index = text_index
160
146
self.signature_index = signature_index
161
self.chk_index = chk_index
163
148
def access_tuple(self):
164
149
"""Return a tuple (transport, name) for the pack content."""
165
150
return self.pack_transport, self.file_name()
167
def _check_references(self):
168
"""Make sure our external references are present.
170
Packs are allowed to have deltas whose base is not in the pack, but it
171
must be present somewhere in this collection. It is not allowed to
172
have deltas based on a fallback repository.
173
(See <https://bugs.launchpad.net/bzr/+bug/288751>)
176
for (index_name, external_refs, index) in [
178
self._get_external_refs(self.text_index),
179
self._pack_collection.text_index.combined_index),
181
self._get_external_refs(self.inventory_index),
182
self._pack_collection.inventory_index.combined_index),
184
missing = external_refs.difference(
185
k for (idx, k, v, r) in
186
index.iter_entries(external_refs))
188
missing_items[index_name] = sorted(list(missing))
190
from pprint import pformat
191
raise errors.BzrCheckError(
192
"Newly created pack file %r has delta references to "
193
"items not in its repository:\n%s"
194
% (self, pformat(missing_items)))
196
152
def file_name(self):
197
153
"""Get the file name for the pack on disk."""
198
154
return self.name + '.pack'
224
172
"""The text index is the name + .tix."""
225
173
return self.index_name('text', name)
227
def _replace_index_with_readonly(self, index_type):
228
unlimited_cache = False
229
if index_type == 'chk':
230
unlimited_cache = True
231
setattr(self, index_type + '_index',
232
self.index_class(self.index_transport,
233
self.index_name(index_type, self.name),
234
self.index_sizes[self.index_offset(index_type)],
235
unlimited_cache=unlimited_cache))
175
def _external_compression_parents_of_texts(self):
178
for node in self.text_index.iter_all_entries():
180
refs.update(node[3][1])
238
184
class ExistingPack(Pack):
239
185
"""An in memory proxy for an existing .pack and its disk indices."""
241
187
def __init__(self, pack_transport, name, revision_index, inventory_index,
242
text_index, signature_index, chk_index=None):
188
text_index, signature_index):
243
189
"""Create an ExistingPack object.
245
191
:param pack_transport: The transport where the pack file resides.
246
192
:param name: The name of the pack on disk in the pack_transport.
248
194
Pack.__init__(self, revision_index, inventory_index, text_index,
249
signature_index, chk_index)
251
197
self.pack_transport = pack_transport
252
198
if None in (revision_index, inventory_index, text_index,
260
206
return not self.__eq__(other)
262
208
def __repr__(self):
263
return "<%s.%s object at 0x%x, %s, %s" % (
264
self.__class__.__module__, self.__class__.__name__, id(self),
265
self.pack_transport, self.name)
268
class ResumedPack(ExistingPack):
270
def __init__(self, name, revision_index, inventory_index, text_index,
271
signature_index, upload_transport, pack_transport, index_transport,
272
pack_collection, chk_index=None):
273
"""Create a ResumedPack object."""
274
ExistingPack.__init__(self, pack_transport, name, revision_index,
275
inventory_index, text_index, signature_index,
277
self.upload_transport = upload_transport
278
self.index_transport = index_transport
279
self.index_sizes = [None, None, None, None]
281
('revision', revision_index),
282
('inventory', inventory_index),
283
('text', text_index),
284
('signature', signature_index),
286
if chk_index is not None:
287
indices.append(('chk', chk_index))
288
self.index_sizes.append(None)
289
for index_type, index in indices:
290
offset = self.index_offset(index_type)
291
self.index_sizes[offset] = index._size
292
self.index_class = pack_collection._index_class
293
self._pack_collection = pack_collection
294
self._state = 'resumed'
295
# XXX: perhaps check that the .pack file exists?
297
def access_tuple(self):
298
if self._state == 'finished':
299
return Pack.access_tuple(self)
300
elif self._state == 'resumed':
301
return self.upload_transport, self.file_name()
303
raise AssertionError(self._state)
306
self.upload_transport.delete(self.file_name())
307
indices = [self.revision_index, self.inventory_index, self.text_index,
308
self.signature_index]
309
if self.chk_index is not None:
310
indices.append(self.chk_index)
311
for index in indices:
312
index._transport.delete(index._name)
315
self._check_references()
316
index_types = ['revision', 'inventory', 'text', 'signature']
317
if self.chk_index is not None:
318
index_types.append('chk')
319
for index_type in index_types:
320
old_name = self.index_name(index_type, self.name)
321
new_name = '../indices/' + old_name
322
self.upload_transport.rename(old_name, new_name)
323
self._replace_index_with_readonly(index_type)
324
new_name = '../packs/' + self.file_name()
325
self.upload_transport.rename(self.file_name(), new_name)
326
self._state = 'finished'
328
def _get_external_refs(self, index):
329
"""Return compression parents for this index that are not present.
331
This returns any compression parents that are referenced by this index,
332
which are not contained *in* this index. They may be present elsewhere.
334
return index.external_references(1)
209
return "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (
210
id(self), self.pack_transport, self.name)
337
213
class NewPack(Pack):
338
214
"""An in memory proxy for a pack which is being created."""
340
def __init__(self, pack_collection, upload_suffix='', file_mode=None):
216
# A map of index 'type' to the file extension and position in the
218
index_definitions = {
219
'revision': ('.rix', 0),
220
'inventory': ('.iix', 1),
222
'signature': ('.six', 3),
225
def __init__(self, upload_transport, index_transport, pack_transport,
226
upload_suffix='', file_mode=None, index_builder_class=None,
341
228
"""Create a NewPack instance.
343
:param pack_collection: A PackCollection into which this is being inserted.
230
:param upload_transport: A writable transport for the pack to be
231
incrementally uploaded to.
232
:param index_transport: A writable transport for the pack's indices to
233
be written to when the pack is finished.
234
:param pack_transport: A writable transport for the pack to be renamed
235
to when the upload is complete. This *must* be the same as
236
upload_transport.clone('../packs').
344
237
:param upload_suffix: An optional suffix to be given to any temporary
345
238
files created during the pack creation. e.g '.autopack'
346
:param file_mode: Unix permissions for newly created file.
239
:param file_mode: An optional file mode to create the new files with.
240
:param index_builder_class: Required keyword parameter - the class of
241
index builder to use.
242
:param index_class: Required keyword parameter - the class of index
348
245
# The relative locations of the packs are constrained, but all are
349
246
# passed in because the caller has them, so as to avoid object churn.
350
index_builder_class = pack_collection._index_builder_class
351
if pack_collection.chk_index is not None:
352
chk_index = index_builder_class(reference_lists=0)
355
247
Pack.__init__(self,
356
248
# Revisions: parents list, no text compression.
357
249
index_builder_class(reference_lists=1),
366
258
# Signatures: Just blobs to store, no compression, no parents
368
260
index_builder_class(reference_lists=0),
369
# CHK based storage - just blobs, no compression or parents.
372
self._pack_collection = pack_collection
373
262
# When we make readonly indices, we need this.
374
self.index_class = pack_collection._index_class
263
self.index_class = index_class
375
264
# where should the new pack be opened
376
self.upload_transport = pack_collection._upload_transport
265
self.upload_transport = upload_transport
377
266
# where are indices written out to
378
self.index_transport = pack_collection._index_transport
267
self.index_transport = index_transport
379
268
# where is the pack renamed to when it is finished?
380
self.pack_transport = pack_collection._pack_transport
269
self.pack_transport = pack_transport
381
270
# What file mode to upload the pack and indices with.
382
271
self._file_mode = file_mode
383
272
# tracks the content written to the .pack file.
384
273
self._hash = osutils.md5()
385
# a tuple with the length in bytes of the indices, once the pack
386
# is finalised. (rev, inv, text, sigs, chk_if_in_use)
274
# a four-tuple with the length in bytes of the indices, once the pack
275
# is finalised. (rev, inv, text, sigs)
387
276
self.index_sizes = None
388
277
# How much data to cache when writing packs. Note that this is not
389
278
# synchronised with reads, because it's not in the transport layer, so
474
352
- stores the index size tuple for the pack in the index_sizes
477
self.finish_content()
479
self._check_references()
357
self._write_data('', flush=True)
358
self.name = self._hash.hexdigest()
481
360
# XXX: It'd be better to write them all to temporary names, then
482
361
# rename them all into place, so that the window when only some are
483
362
# visible is smaller. On the other hand none will be seen until
484
363
# they're in the names list.
485
364
self.index_sizes = [None, None, None, None]
486
self._write_index('revision', self.revision_index, 'revision', suspend)
487
self._write_index('inventory', self.inventory_index, 'inventory',
489
self._write_index('text', self.text_index, 'file texts', suspend)
365
self._write_index('revision', self.revision_index, 'revision')
366
self._write_index('inventory', self.inventory_index, 'inventory')
367
self._write_index('text', self.text_index, 'file texts')
490
368
self._write_index('signature', self.signature_index,
491
'revision signatures', suspend)
492
if self.chk_index is not None:
493
self.index_sizes.append(None)
494
self._write_index('chk', self.chk_index,
495
'content hash bytes', suspend)
369
'revision signatures')
496
370
self.write_stream.close()
497
371
# Note that this will clobber an existing pack with the same name,
498
372
# without checking for hash collisions. While this is undesirable this
524
397
self._hash.update(bytes)
525
398
self._buffer[:] = [[], 0]
527
def _get_external_refs(self, index):
528
return index._external_references()
400
def index_name(self, index_type, name):
401
"""Get the disk name of an index type for pack name 'name'."""
402
return name + NewPack.index_definitions[index_type][0]
404
def index_offset(self, index_type):
405
"""Get the position in a index_size array for a given index type."""
406
return NewPack.index_definitions[index_type][1]
408
def _replace_index_with_readonly(self, index_type):
409
setattr(self, index_type + '_index',
410
self.index_class(self.index_transport,
411
self.index_name(index_type, self.name),
412
self.index_sizes[self.index_offset(index_type)]))
530
414
def set_write_cache_size(self, size):
531
415
self._cache_limit = size
533
def _write_index(self, index_type, index, label, suspend=False):
417
def _write_index(self, index_type, index, label):
534
418
"""Write out an index.
536
420
:param index_type: The type of index to write - e.g. 'revision'.
538
422
:param label: What label to give the index e.g. 'revision'.
540
424
index_name = self.index_name(index_type, self.name)
542
transport = self.upload_transport
544
transport = self.index_transport
545
self.index_sizes[self.index_offset(index_type)] = transport.put_file(
546
index_name, index.finish(), mode=self._file_mode)
425
self.index_sizes[self.index_offset(index_type)] = \
426
self.index_transport.put_file(index_name, index.finish(),
427
mode=self._file_mode)
547
428
if 'pack' in debug.debug_flags:
548
429
# XXX: size might be interesting?
549
430
mutter('%s: create_pack: wrote %s index: %s%s t+%6.3fs',
550
431
time.ctime(), label, self.upload_transport.base,
551
432
self.random_name, time.time() - self.start_time)
552
# Replace the writable index on this object with a readonly,
433
# Replace the writable index on this object with a readonly,
553
434
# presently unloaded index. We should alter
554
435
# the index layer to make its finish() error if add_node is
555
436
# subsequently used. RBC
564
445
such as 'revision index'.
566
447
A CombinedIndex provides an index on a single key space built up
567
from several on-disk indices. The AggregateIndex builds on this
448
from several on-disk indices. The AggregateIndex builds on this
568
449
to provide a knit access layer, and allows having up to one writable
569
450
index within the collection.
571
452
# XXX: Probably 'can be written to' could/should be separated from 'acts
572
453
# like a knit index' -- mbp 20071024
574
def __init__(self, reload_func=None, flush_func=None):
575
"""Create an AggregateIndex.
577
:param reload_func: A function to call if we find we are missing an
578
index. Should have the form reload_func() => True if the list of
579
active pack files has changed.
581
self._reload_func = reload_func
456
"""Create an AggregateIndex."""
582
457
self.index_to_pack = {}
583
self.combined_index = CombinedGraphIndex([], reload_func=reload_func)
584
self.data_access = _DirectPackAccess(self.index_to_pack,
585
reload_func=reload_func,
586
flush_func=flush_func)
458
self.combined_index = CombinedGraphIndex([])
459
self.data_access = _DirectPackAccess(self.index_to_pack)
587
460
self.add_callback = None
589
462
def replace_indices(self, index_to_pack, indices):
694
562
def _extra_init(self):
695
563
"""A template hook to allow extending the constructor trivially."""
697
def _pack_map_and_index_list(self, index_attribute):
698
"""Convert a list of packs to an index pack map and index list.
700
:param index_attribute: The attribute that the desired index is found
702
:return: A tuple (map, list) where map contains the dict from
703
index:pack_tuple, and list contains the indices in the preferred
708
for pack_obj in self.packs:
709
index = getattr(pack_obj, index_attribute)
710
indices.append(index)
711
pack_map[index] = pack_obj
712
return pack_map, indices
714
def _index_contents(self, indices, key_filter=None):
715
"""Get an iterable of the index contents from a pack_map.
717
:param indices: The list of indices to query
718
:param key_filter: An optional filter to limit the keys returned.
720
all_index = CombinedGraphIndex(indices)
721
if key_filter is None:
722
return all_index.iter_all_entries()
724
return all_index.iter_entries(key_filter)
726
565
def pack(self, pb=None):
727
566
"""Create a new pack by reading data from other packs.
729
568
This does little more than a bulk copy of data. One key difference
730
569
is that data with the same item key across multiple packs is elided
731
570
from the output. The new pack is written into the current pack store
732
along with its indices, and the name added to the pack names. The
571
along with its indices, and the name added to the pack names. The
733
572
source packs are not altered and are not required to be in the current
766
604
def open_pack(self):
767
605
"""Open a pack for the pack we are creating."""
768
new_pack = self._pack_collection.pack_factory(self._pack_collection,
769
upload_suffix=self.suffix,
770
file_mode=self._pack_collection.repo.bzrdir._get_file_mode())
771
# We know that we will process all nodes in order, and don't need to
772
# query, so don't combine any indices spilled to disk until we are done
773
new_pack.revision_index.set_optimize(combine_backing_indices=False)
774
new_pack.inventory_index.set_optimize(combine_backing_indices=False)
775
new_pack.text_index.set_optimize(combine_backing_indices=False)
776
new_pack.signature_index.set_optimize(combine_backing_indices=False)
779
def _update_pack_order(self, entries, index_to_pack_map):
780
"""Determine how we want our packs to be ordered.
782
This changes the sort order of the self.packs list so that packs unused
783
by 'entries' will be at the end of the list, so that future requests
784
can avoid probing them. Used packs will be at the front of the
785
self.packs list, in the order of their first use in 'entries'.
787
:param entries: A list of (index, ...) tuples
788
:param index_to_pack_map: A mapping from index objects to pack objects.
792
for entry in entries:
794
if index not in seen_indexes:
795
packs.append(index_to_pack_map[index])
796
seen_indexes.add(index)
797
if len(packs) == len(self.packs):
798
if 'pack' in debug.debug_flags:
799
mutter('Not changing pack list, all packs used.')
801
seen_packs = set(packs)
802
for pack in self.packs:
803
if pack not in seen_packs:
806
if 'pack' in debug.debug_flags:
807
old_names = [p.access_tuple()[1] for p in self.packs]
808
new_names = [p.access_tuple()[1] for p in packs]
809
mutter('Reordering packs\nfrom: %s\n to: %s',
810
old_names, new_names)
606
return NewPack(self._pack_collection._upload_transport,
607
self._pack_collection._index_transport,
608
self._pack_collection._pack_transport, upload_suffix=self.suffix,
609
file_mode=self._pack_collection.repo.bzrdir._get_file_mode(),
610
index_builder_class=self._pack_collection._index_builder_class,
611
index_class=self._pack_collection._index_class)
813
613
def _copy_revision_texts(self):
814
614
"""Copy revision data to the new pack."""
957
756
self._pack_collection.allocate(new_pack)
960
def _copy_chks(self, refs=None):
961
# XXX: Todo, recursive follow-pointers facility when fetching some
963
chk_index_map, chk_indices = self._pack_map_and_index_list(
965
chk_nodes = self._index_contents(chk_indices, refs)
967
# TODO: This isn't strictly tasteful as we are accessing some private
968
# variables (_serializer). Perhaps a better way would be to have
969
# Repository._deserialise_chk_node()
970
search_key_func = chk_map.search_key_registry.get(
971
self._pack_collection.repo._serializer.search_key_name)
972
def accumlate_refs(lines):
973
# XXX: move to a generic location
975
bytes = ''.join(lines)
976
node = chk_map._deserialise(bytes, ("unknown",), search_key_func)
977
new_refs.update(node.refs())
978
self._copy_nodes(chk_nodes, chk_index_map, self.new_pack._writer,
979
self.new_pack.chk_index, output_lines=accumlate_refs)
982
def _copy_nodes(self, nodes, index_map, writer, write_index,
984
"""Copy knit nodes between packs with no graph references.
986
:param output_lines: Output full texts of copied items.
759
def _copy_nodes(self, nodes, index_map, writer, write_index):
760
"""Copy knit nodes between packs with no graph references."""
988
761
pb = ui.ui_factory.nested_progress_bar()
990
763
return self._do_copy_nodes(nodes, index_map, writer,
991
write_index, pb, output_lines=output_lines)
995
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb,
768
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
997
769
# for record verification
998
770
knit = KnitVersionedFiles(None, None)
999
771
# plan a readv on each source pack:
1020
792
# linear scan up the pack
1021
793
pack_readv_requests.sort()
1023
pack_obj = index_map[index]
1024
transport, path = pack_obj.access_tuple()
1026
reader = pack.make_readv_reader(transport, path,
1027
[offset[0:2] for offset in pack_readv_requests])
1028
except errors.NoSuchFile:
1029
if self._reload_func is not None:
795
transport, path = index_map[index]
796
reader = pack.make_readv_reader(transport, path,
797
[offset[0:2] for offset in pack_readv_requests])
1032
798
for (names, read_func), (_1, _2, (key, eol_flag)) in \
1033
799
izip(reader.iter_records(), pack_readv_requests):
1034
800
raw_data = read_func(None)
1035
801
# check the header only
1036
if output_lines is not None:
1037
output_lines(knit._parse_record(key[-1], raw_data)[0])
1039
df, _ = knit._parse_record_header(key, raw_data)
802
df, _ = knit._parse_record_header(key, raw_data)
1041
804
pos, size = writer.add_bytes_record(raw_data, names)
1042
805
write_index.add_node(key, eol_flag + "%d %d" % (pos, size))
1043
806
pb.update("Copied record", record_index)
1103
860
record_index += 1
1105
862
def _get_text_nodes(self):
1106
text_index_map, text_indices = self._pack_map_and_index_list(
1108
return text_index_map, self._index_contents(text_indices,
863
text_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
864
self.packs, 'text_index')[0]
865
return text_index_map, self._pack_collection._index_contents(text_index_map,
1109
866
self._text_filter)
1111
868
def _least_readv_node_readv(self, nodes):
1112
869
"""Generate request groups for nodes using the least readv's.
1114
871
:param nodes: An iterable of graph index nodes.
1115
872
:return: Total node count and an iterator of the data needed to perform
1116
873
readvs to obtain the data for nodes. Each item yielded by the
1117
874
iterator is a tuple with:
1118
875
index, readv_vector, node_vector. readv_vector is a list ready to
1119
876
hand to the transport readv method, and node_vector is a list of
1120
(key, eol_flag, references) for the node retrieved by the
877
(key, eol_flag, references) for the the node retrieved by the
1121
878
matching readv_vector.
1123
880
# group by pack so we do one readv per pack
1399
1140
self._pack_transport = pack_transport
1400
1141
self._index_builder_class = index_builder_class
1401
1142
self._index_class = index_class
1402
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3,
1143
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1404
1144
self.packs = []
1405
1145
# name:Pack mapping
1407
1146
self._packs_by_name = {}
1408
1147
# the previous pack-names content
1409
1148
self._packs_at_load = None
1410
1149
# when a pack is being created by this object, the state of that pack.
1411
1150
self._new_pack = None
1412
1151
# aggregated revision index data
1413
flush = self._flush_new_pack
1414
self.revision_index = AggregateIndex(self.reload_pack_names, flush)
1415
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1416
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1417
self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1419
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1421
# used to determine if we're using a chk_index elsewhere.
1422
self.chk_index = None
1424
self._resumed_packs = []
1427
return '%s(%r)' % (self.__class__.__name__, self.repo)
1152
self.revision_index = AggregateIndex()
1153
self.inventory_index = AggregateIndex()
1154
self.text_index = AggregateIndex()
1155
self.signature_index = AggregateIndex()
1429
1157
def add_pack_to_memory(self, pack):
1430
1158
"""Make a Pack object available to the repository to satisfy queries.
1432
1160
:param pack: A Pack object.
1434
1162
if pack.name in self._packs_by_name:
1435
raise AssertionError(
1436
'pack %s already in _packs_by_name' % (pack.name,))
1163
raise AssertionError()
1437
1164
self.packs.append(pack)
1438
1165
self._packs_by_name[pack.name] = pack
1439
1166
self.revision_index.add_index(pack.revision_index, pack)
1440
1167
self.inventory_index.add_index(pack.inventory_index, pack)
1441
1168
self.text_index.add_index(pack.text_index, pack)
1442
1169
self.signature_index.add_index(pack.signature_index, pack)
1443
if self.chk_index is not None:
1444
self.chk_index.add_index(pack.chk_index, pack)
1446
1171
def all_packs(self):
1447
1172
"""Return a list of all the Pack objects this repository has.
1469
1194
in synchronisation with certain steps. Otherwise the names collection
1470
1195
is not flushed.
1472
:return: Something evaluating true if packing took place.
1197
:return: True if packing took place.
1476
return self._do_autopack()
1477
except errors.RetryAutopack:
1478
# If we get a RetryAutopack exception, we should abort the
1479
# current action, and retry.
1482
def _do_autopack(self):
1483
1199
# XXX: Should not be needed when the management of indices is sane.
1484
1200
total_revisions = self.revision_index.combined_index.key_count()
1485
1201
total_packs = len(self._names)
1486
1202
if self._max_pack_count(total_revisions) >= total_packs:
1204
# XXX: the following may want to be a class, to pack with a given
1206
mutter('Auto-packing repository %s, which has %d pack files, '
1207
'containing %d revisions into %d packs.', self, total_packs,
1208
total_revisions, self._max_pack_count(total_revisions))
1488
1209
# determine which packs need changing
1489
1210
pack_distribution = self.pack_distribution(total_revisions)
1490
1211
existing_packs = []
1498
1219
# group their data with the relevant commit, and that may
1499
1220
# involve rewriting ancient history - which autopack tries to
1500
1221
# avoid. Alternatively we could not group the data but treat
1501
# each of these as having a single revision, and thus add
1222
# each of these as having a single revision, and thus add
1502
1223
# one revision for each to the total revision count, to get
1503
1224
# a matching distribution.
1505
1226
existing_packs.append((revision_count, pack))
1506
1227
pack_operations = self.plan_autopack_combinations(
1507
1228
existing_packs, pack_distribution)
1508
num_new_packs = len(pack_operations)
1509
num_old_packs = sum([len(po[1]) for po in pack_operations])
1510
num_revs_affected = sum([po[0] for po in pack_operations])
1511
mutter('Auto-packing repository %s, which has %d pack files, '
1512
'containing %d revisions. Packing %d files into %d affecting %d'
1513
' revisions', self, total_packs, total_revisions, num_old_packs,
1514
num_new_packs, num_revs_affected)
1515
result = self._execute_pack_operations(pack_operations,
1516
reload_func=self._restart_autopack)
1517
mutter('Auto-packing repository %s completed', self)
1229
self._execute_pack_operations(pack_operations)
1520
def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1232
def _execute_pack_operations(self, pack_operations, _packer_class=Packer):
1522
1233
"""Execute a series of pack operations.
1524
1235
:param pack_operations: A list of [revision_count, packs_to_combine].
1525
1236
:param _packer_class: The class of packer to use (default: Packer).
1526
:return: The new pack names.
1528
1239
for revision_count, packs in pack_operations:
1529
1240
# we may have no-ops from the setup logic
1530
1241
if len(packs) == 0:
1532
packer = _packer_class(self, packs, '.autopack',
1533
reload_func=reload_func)
1536
except errors.RetryWithNewPacks:
1537
# An exception is propagating out of this context, make sure
1538
# this packer has cleaned up. Packer() doesn't set its new_pack
1539
# state into the RepositoryPackCollection object, so we only
1540
# have access to it directly here.
1541
if packer.new_pack is not None:
1542
packer.new_pack.abort()
1243
_packer_class(self, packs, '.autopack').pack()
1544
1244
for pack in packs:
1545
1245
self._remove_pack_from_memory(pack)
1546
1246
# record the newly available packs and stop advertising the old
1548
to_be_obsoleted = []
1549
for _, packs in pack_operations:
1550
to_be_obsoleted.extend(packs)
1551
result = self._save_pack_names(clear_obsolete_packs=True,
1552
obsolete_packs=to_be_obsoleted)
1555
def _flush_new_pack(self):
1556
if self._new_pack is not None:
1557
self._new_pack.flush()
1248
self._save_pack_names(clear_obsolete_packs=True)
1249
# Move the old packs out of the way now they are no longer referenced.
1250
for revision_count, packs in pack_operations:
1251
self._obsolete_packs(packs)
1559
1253
def lock_names(self):
1560
1254
"""Acquire the mutex around the pack-names index.
1562
1256
This cannot be used in the middle of a read-only transaction on the
1565
1259
self.repo.control_files.lock_write()
1567
def _already_packed(self):
1568
"""Is the collection already packed?"""
1569
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1571
def pack(self, hint=None):
1572
1262
"""Pack the pack collection totally."""
1573
1263
self.ensure_loaded()
1574
1264
total_packs = len(self._names)
1575
if self._already_packed():
1266
# This is arguably wrong because we might not be optimal, but for
1267
# now lets leave it in. (e.g. reconcile -> one pack. But not
1577
1270
total_revisions = self.revision_index.combined_index.key_count()
1578
1271
# XXX: the following may want to be a class, to pack with a given
1580
1273
mutter('Packing repository %s, which has %d pack files, '
1581
'containing %d revisions with hint %r.', self, total_packs,
1582
total_revisions, hint)
1274
'containing %d revisions into 1 packs.', self, total_packs,
1583
1276
# determine which packs need changing
1277
pack_distribution = [1]
1584
1278
pack_operations = [[0, []]]
1585
1279
for pack in self.all_packs():
1586
if hint is None or pack.name in hint:
1587
# Either no hint was provided (so we are packing everything),
1588
# or this pack was included in the hint.
1589
pack_operations[-1][0] += pack.get_revision_count()
1590
pack_operations[-1][1].append(pack)
1280
pack_operations[-1][0] += pack.get_revision_count()
1281
pack_operations[-1][1].append(pack)
1591
1282
self._execute_pack_operations(pack_operations, OptimisingPacker)
1593
1284
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1682
1365
inv_index = self._make_index(name, '.iix')
1683
1366
txt_index = self._make_index(name, '.tix')
1684
1367
sig_index = self._make_index(name, '.six')
1685
if self.chk_index is not None:
1686
chk_index = self._make_index(name, '.cix', unlimited_cache=True)
1689
1368
result = ExistingPack(self._pack_transport, name, rev_index,
1690
inv_index, txt_index, sig_index, chk_index)
1369
inv_index, txt_index, sig_index)
1691
1370
self.add_pack_to_memory(result)
1694
def _resume_pack(self, name):
1695
"""Get a suspended Pack object by name.
1697
:param name: The name of the pack - e.g. '123456'
1698
:return: A Pack object.
1700
if not re.match('[a-f0-9]{32}', name):
1701
# Tokens should be md5sums of the suspended pack file, i.e. 32 hex
1703
raise errors.UnresumableWriteGroup(
1704
self.repo, [name], 'Malformed write group token')
1706
rev_index = self._make_index(name, '.rix', resume=True)
1707
inv_index = self._make_index(name, '.iix', resume=True)
1708
txt_index = self._make_index(name, '.tix', resume=True)
1709
sig_index = self._make_index(name, '.six', resume=True)
1710
if self.chk_index is not None:
1711
chk_index = self._make_index(name, '.cix', resume=True,
1712
unlimited_cache=True)
1715
result = self.resumed_pack_factory(name, rev_index, inv_index,
1716
txt_index, sig_index, self._upload_transport,
1717
self._pack_transport, self._index_transport, self,
1718
chk_index=chk_index)
1719
except errors.NoSuchFile, e:
1720
raise errors.UnresumableWriteGroup(self.repo, [name], str(e))
1721
self.add_pack_to_memory(result)
1722
self._resumed_packs.append(result)
1725
1373
def allocate(self, a_new_pack):
1726
1374
"""Allocate name in the list of packs.
1745
1393
return self._index_class(self.transport, 'pack-names', None
1746
1394
).iter_all_entries()
1748
def _make_index(self, name, suffix, resume=False, unlimited_cache=False):
1396
def _make_index(self, name, suffix):
1749
1397
size_offset = self._suffix_offsets[suffix]
1750
1398
index_name = name + suffix
1752
transport = self._upload_transport
1753
index_size = transport.stat(index_name).st_size
1755
transport = self._index_transport
1756
index_size = self._names[name][size_offset]
1757
return self._index_class(transport, index_name, index_size,
1758
unlimited_cache=unlimited_cache)
1399
index_size = self._names[name][size_offset]
1400
return self._index_class(
1401
self._index_transport, index_name, index_size)
1760
1403
def _max_pack_count(self, total_revisions):
1761
1404
"""Return the maximum number of packs to use for total revisions.
1763
1406
:param total_revisions: The total number of revisions in the
1789
1432
:param return: None.
1791
1434
for pack in packs:
1793
pack.pack_transport.rename(pack.file_name(),
1794
'../obsolete_packs/' + pack.file_name())
1795
except (errors.PathError, errors.TransportError), e:
1796
# TODO: Should these be warnings or mutters?
1797
mutter("couldn't rename obsolete pack, skipping it:\n%s"
1435
pack.pack_transport.rename(pack.file_name(),
1436
'../obsolete_packs/' + pack.file_name())
1799
1437
# TODO: Probably needs to know all possible indices for this pack
1800
1438
# - or maybe list the directory and move all indices matching this
1801
1439
# name whether we recognize it or not?
1802
suffixes = ['.iix', '.six', '.tix', '.rix']
1803
if self.chk_index is not None:
1804
suffixes.append('.cix')
1805
for suffix in suffixes:
1807
self._index_transport.rename(pack.name + suffix,
1808
'../obsolete_packs/' + pack.name + suffix)
1809
except (errors.PathError, errors.TransportError), e:
1810
mutter("couldn't rename obsolete index, skipping it:\n%s"
1440
for suffix in ('.iix', '.six', '.tix', '.rix'):
1441
self._index_transport.rename(pack.name + suffix,
1442
'../obsolete_packs/' + pack.name + suffix)
1813
1444
def pack_distribution(self, total_revisions):
1814
1445
"""Generate a list of the number of revisions to put in each pack.
1870
1498
self._packs_by_name = {}
1871
1499
self._packs_at_load = None
1501
def _make_index_map(self, index_suffix):
1502
"""Return information on existing indices.
1504
:param suffix: Index suffix added to pack name.
1506
:returns: (pack_map, indices) where indices is a list of GraphIndex
1507
objects, and pack_map is a mapping from those objects to the
1508
pack tuple they describe.
1510
# TODO: stop using this; it creates new indices unnecessarily.
1511
self.ensure_loaded()
1512
suffix_map = {'.rix': 'revision_index',
1513
'.six': 'signature_index',
1514
'.iix': 'inventory_index',
1515
'.tix': 'text_index',
1517
return self._packs_list_to_pack_map_and_index_list(self.all_packs(),
1518
suffix_map[index_suffix])
1520
def _packs_list_to_pack_map_and_index_list(self, packs, index_attribute):
1521
"""Convert a list of packs to an index pack map and index list.
1523
:param packs: The packs list to process.
1524
:param index_attribute: The attribute that the desired index is found
1526
:return: A tuple (map, list) where map contains the dict from
1527
index:pack_tuple, and lsit contains the indices in the same order
1533
index = getattr(pack, index_attribute)
1534
indices.append(index)
1535
pack_map[index] = (pack.pack_transport, pack.file_name())
1536
return pack_map, indices
1538
def _index_contents(self, pack_map, key_filter=None):
1539
"""Get an iterable of the index contents from a pack_map.
1541
:param pack_map: A map from indices to pack details.
1542
:param key_filter: An optional filter to limit the
1545
indices = [index for index in pack_map.iterkeys()]
1546
all_index = CombinedGraphIndex(indices)
1547
if key_filter is None:
1548
return all_index.iter_all_entries()
1550
return all_index.iter_entries(key_filter)
1873
1552
def _unlock_names(self):
1874
1553
"""Release the mutex around the pack-names index."""
1875
1554
self.repo.control_files.unlock()
1877
def _diff_pack_names(self):
1878
"""Read the pack names from disk, and compare it to the one in memory.
1880
:return: (disk_nodes, deleted_nodes, new_nodes)
1881
disk_nodes The final set of nodes that should be referenced
1882
deleted_nodes Nodes which have been removed from when we started
1883
new_nodes Nodes that are newly introduced
1885
# load the disk nodes across
1887
for index, key, value in self._iter_disk_pack_index():
1888
disk_nodes.add((key, value))
1889
orig_disk_nodes = set(disk_nodes)
1891
# do a two-way diff against our original content
1892
current_nodes = set()
1893
for name, sizes in self._names.iteritems():
1895
((name, ), ' '.join(str(size) for size in sizes)))
1897
# Packs no longer present in the repository, which were present when we
1898
# locked the repository
1899
deleted_nodes = self._packs_at_load - current_nodes
1900
# Packs which this process is adding
1901
new_nodes = current_nodes - self._packs_at_load
1903
# Update the disk_nodes set to include the ones we are adding, and
1904
# remove the ones which were removed by someone else
1905
disk_nodes.difference_update(deleted_nodes)
1906
disk_nodes.update(new_nodes)
1908
return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
1910
def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1911
"""Given the correct set of pack files, update our saved info.
1913
:return: (removed, added, modified)
1914
removed pack names removed from self._names
1915
added pack names added to self._names
1916
modified pack names that had changed value
1921
## self._packs_at_load = disk_nodes
1556
def _save_pack_names(self, clear_obsolete_packs=False):
1557
"""Save the list of packs.
1559
This will take out the mutex around the pack names list for the
1560
duration of the method call. If concurrent updates have been made, a
1561
three-way merge between the current list and the current in memory list
1564
:param clear_obsolete_packs: If True, clear out the contents of the
1565
obsolete_packs directory.
1569
builder = self._index_builder_class()
1570
# load the disk nodes across
1572
for index, key, value in self._iter_disk_pack_index():
1573
disk_nodes.add((key, value))
1574
# do a two-way diff against our original content
1575
current_nodes = set()
1576
for name, sizes in self._names.iteritems():
1578
((name, ), ' '.join(str(size) for size in sizes)))
1579
deleted_nodes = self._packs_at_load - current_nodes
1580
new_nodes = current_nodes - self._packs_at_load
1581
disk_nodes.difference_update(deleted_nodes)
1582
disk_nodes.update(new_nodes)
1583
# TODO: handle same-name, index-size-changes here -
1584
# e.g. use the value from disk, not ours, *unless* we're the one
1586
for key, value in disk_nodes:
1587
builder.add_node(key, value)
1588
self.transport.put_file('pack-names', builder.finish(),
1589
mode=self.repo.bzrdir._get_file_mode())
1590
# move the baseline forward
1591
self._packs_at_load = disk_nodes
1592
if clear_obsolete_packs:
1593
self._clear_obsolete_packs()
1595
self._unlock_names()
1596
# synchronise the memory packs list with what we just wrote:
1922
1597
new_names = dict(disk_nodes)
1923
1598
# drop no longer present nodes
1924
1599
for pack in self.all_packs():
1925
1600
if (pack.name,) not in new_names:
1926
removed.append(pack.name)
1927
1601
self._remove_pack_from_memory(pack)
1928
1602
# add new nodes/refresh existing ones
1929
1603
for key, value in disk_nodes:
1943
1617
self._remove_pack_from_memory(self.get_pack_by_name(name))
1944
1618
self._names[name] = sizes
1945
1619
self.get_pack_by_name(name)
1946
modified.append(name)
1949
1622
self._names[name] = sizes
1950
1623
self.get_pack_by_name(name)
1952
return removed, added, modified
1954
def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
1955
"""Save the list of packs.
1957
This will take out the mutex around the pack names list for the
1958
duration of the method call. If concurrent updates have been made, a
1959
three-way merge between the current list and the current in memory list
1962
:param clear_obsolete_packs: If True, clear out the contents of the
1963
obsolete_packs directory.
1964
:param obsolete_packs: Packs that are obsolete once the new pack-names
1965
file has been written.
1966
:return: A list of the names saved that were not previously on disk.
1968
already_obsolete = []
1971
builder = self._index_builder_class()
1972
(disk_nodes, deleted_nodes, new_nodes,
1973
orig_disk_nodes) = self._diff_pack_names()
1974
# TODO: handle same-name, index-size-changes here -
1975
# e.g. use the value from disk, not ours, *unless* we're the one
1977
for key, value in disk_nodes:
1978
builder.add_node(key, value)
1979
self.transport.put_file('pack-names', builder.finish(),
1980
mode=self.repo.bzrdir._get_file_mode())
1981
self._packs_at_load = disk_nodes
1982
if clear_obsolete_packs:
1985
to_preserve = set([o.name for o in obsolete_packs])
1986
already_obsolete = self._clear_obsolete_packs(to_preserve)
1988
self._unlock_names()
1989
# synchronise the memory packs list with what we just wrote:
1990
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1992
# TODO: We could add one more condition here. "if o.name not in
1993
# orig_disk_nodes and o != the new_pack we haven't written to
1994
# disk yet. However, the new pack object is not easily
1995
# accessible here (it would have to be passed through the
1996
# autopacking code, etc.)
1997
obsolete_packs = [o for o in obsolete_packs
1998
if o.name not in already_obsolete]
1999
self._obsolete_packs(obsolete_packs)
2000
return [new_node[0][0] for new_node in new_nodes]
2002
def reload_pack_names(self):
2003
"""Sync our pack listing with what is present in the repository.
2005
This should be called when we find out that something we thought was
2006
present is now missing. This happens when another process re-packs the
2009
:return: True if the in-memory list of packs has been altered at all.
2011
# The ensure_loaded call is to handle the case where the first call
2012
# made involving the collection was to reload_pack_names, where we
2013
# don't have a view of disk contents. Its a bit of a bandaid, and
2014
# causes two reads of pack-names, but its a rare corner case not struck
2015
# with regular push/pull etc.
2016
first_read = self.ensure_loaded()
2019
# out the new value.
2020
(disk_nodes, deleted_nodes, new_nodes,
2021
orig_disk_nodes) = self._diff_pack_names()
2022
# _packs_at_load is meant to be the explicit list of names in
2023
# 'pack-names' at then start. As such, it should not contain any
2024
# pending names that haven't been written out yet.
2025
self._packs_at_load = orig_disk_nodes
2027
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2028
if removed or added or modified:
2032
def _restart_autopack(self):
2033
"""Reload the pack names list, and restart the autopack code."""
2034
if not self.reload_pack_names():
2035
# Re-raise the original exception, because something went missing
2036
# and a restart didn't find it
2038
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2040
def _clear_obsolete_packs(self, preserve=None):
1625
def _clear_obsolete_packs(self):
2041
1626
"""Delete everything from the obsolete-packs directory.
2043
:return: A list of pack identifiers (the filename without '.pack') that
2044
were found in obsolete_packs.
2047
1628
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2048
if preserve is None:
2050
1629
for filename in obsolete_pack_transport.list_dir('.'):
2051
name, ext = osutils.splitext(filename)
2054
if name in preserve:
2057
1631
obsolete_pack_transport.delete(filename)
2058
1632
except (errors.PathError, errors.TransportError), e:
2059
warning("couldn't delete obsolete pack, skipping it:\n%s"
1633
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2063
1635
def _start_write_group(self):
2064
1636
# Do not permit preparation for writing if we're not in a 'write lock'.
2065
1637
if not self.repo.is_write_locked():
2066
1638
raise errors.NotWriteLocked(self)
2067
self._new_pack = self.pack_factory(self, upload_suffix='.pack',
2068
file_mode=self.repo.bzrdir._get_file_mode())
1639
self._new_pack = NewPack(self._upload_transport, self._index_transport,
1640
self._pack_transport, upload_suffix='.pack',
1641
file_mode=self.repo.bzrdir._get_file_mode(),
1642
index_builder_class=self._index_builder_class,
1643
index_class=self._index_class)
2069
1644
# allow writing: queue writes to a new index
2070
1645
self.revision_index.add_writable_index(self._new_pack.revision_index,
2071
1646
self._new_pack)
2091
1660
# FIXME: just drop the transient index.
2092
1661
# forget what names there are
2093
1662
if self._new_pack is not None:
2095
self._new_pack.abort()
2097
# XXX: If we aborted while in the middle of finishing the write
2098
# group, _remove_pack_indices can fail because the indexes are
2099
# already gone. If they're not there we shouldn't fail in this
2100
# case. -- mbp 20081113
2101
self._remove_pack_indices(self._new_pack)
2102
self._new_pack = None
2103
for resumed_pack in self._resumed_packs:
2105
resumed_pack.abort()
2107
# See comment in previous finally block.
2109
self._remove_pack_indices(resumed_pack)
2112
del self._resumed_packs[:]
2114
def _remove_resumed_pack_indices(self):
2115
for resumed_pack in self._resumed_packs:
2116
self._remove_pack_indices(resumed_pack)
2117
del self._resumed_packs[:]
2119
def _check_new_inventories(self):
2120
"""Detect missing inventories in this write group.
2122
:returns: list of strs, summarising any problems found. If the list is
2123
empty no problems were found.
2125
# The base implementation does no checks. GCRepositoryPackCollection
1663
self._new_pack.abort()
1664
self._remove_pack_indices(self._new_pack)
1665
self._new_pack = None
1666
self.repo._text_knit = None
2129
1668
def _commit_write_group(self):
2131
for prefix, versioned_file in (
2132
('revisions', self.repo.revisions),
2133
('inventories', self.repo.inventories),
2134
('texts', self.repo.texts),
2135
('signatures', self.repo.signatures),
2137
missing = versioned_file.get_missing_compression_parent_keys()
2138
all_missing.update([(prefix,) + key for key in missing])
2140
raise errors.BzrCheckError(
2141
"Repository %s has missing compression parent(s) %r "
2142
% (self.repo, sorted(all_missing)))
2143
problems = self._check_new_inventories()
2145
problems_summary = '\n'.join(problems)
2146
raise errors.BzrCheckError(
2147
"Cannot add revision(s) to repository: " + problems_summary)
2148
1669
self._remove_pack_indices(self._new_pack)
2149
any_new_content = False
2150
1670
if self._new_pack.data_inserted():
2151
1671
# get all the data to disk and read to use
2152
1672
self._new_pack.finish()
2153
1673
self.allocate(self._new_pack)
2154
1674
self._new_pack = None
2155
any_new_content = True
2157
self._new_pack.abort()
2158
self._new_pack = None
2159
for resumed_pack in self._resumed_packs:
2160
# XXX: this is a pretty ugly way to turn the resumed pack into a
2161
# properly committed pack.
2162
self._names[resumed_pack.name] = None
2163
self._remove_pack_from_memory(resumed_pack)
2164
resumed_pack.finish()
2165
self.allocate(resumed_pack)
2166
any_new_content = True
2167
del self._resumed_packs[:]
2169
result = self.autopack()
1675
if not self.autopack():
2171
1676
# when autopack takes no steps, the names list is still
2173
return self._save_pack_names()
2177
def _suspend_write_group(self):
2178
tokens = [pack.name for pack in self._resumed_packs]
2179
self._remove_pack_indices(self._new_pack)
2180
if self._new_pack.data_inserted():
2181
# get all the data to disk and read to use
2182
self._new_pack.finish(suspend=True)
2183
tokens.append(self._new_pack.name)
2184
self._new_pack = None
1678
self._save_pack_names()
2186
1680
self._new_pack.abort()
2187
1681
self._new_pack = None
2188
self._remove_resumed_pack_indices()
2191
def _resume_write_group(self, tokens):
2192
for token in tokens:
2193
self._resume_pack(token)
1682
self.repo._text_knit = None
2196
1685
class KnitPackRepository(KnitRepository):
2197
1686
"""Repository with knit objects stored inside pack containers.
2199
1688
The layering for a KnitPackRepository is:
2201
1690
Graph | HPSS | Repository public layer |
2256
1742
deltas=True, parents=True, is_locked=self.is_locked),
2257
1743
data_access=self._pack_collection.text_index.data_access,
2258
1744
max_delta_chain=200)
2259
if _format.supports_chks:
2260
# No graph, no compression:- references from chks are between
2261
# different objects not temporal versions of the same; and without
2262
# some sort of temporal structure knit compression will just fail.
2263
self.chk_bytes = KnitVersionedFiles(
2264
_KnitGraphIndex(self._pack_collection.chk_index.combined_index,
2265
add_callback=self._pack_collection.chk_index.add_callback,
2266
deltas=False, parents=False, is_locked=self.is_locked),
2267
data_access=self._pack_collection.chk_index.data_access,
2270
self.chk_bytes = None
2271
1745
# True when the repository object is 'write locked' (as opposed to the
2272
# physical lock only taken out around changes to the pack-names list.)
1746
# physical lock only taken out around changes to the pack-names list.)
2273
1747
# Another way to represent this would be a decorator around the control
2274
1748
# files object that presents logical locks as physical ones - if this
2275
1749
# gets ugly consider that alternative design. RBC 20071011
2279
1753
self._reconcile_does_inventory_gc = True
2280
1754
self._reconcile_fixes_text_parents = True
2281
1755
self._reconcile_backsup_inventory = False
1756
self._fetch_order = 'unordered'
2283
def _warn_if_deprecated(self, branch=None):
1758
def _warn_if_deprecated(self):
2284
1759
# This class isn't deprecated, but one sub-format is
2285
1760
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2286
super(KnitPackRepository, self)._warn_if_deprecated(branch)
1761
from bzrlib import repository
1762
if repository._deprecation_warning_done:
1764
repository._deprecation_warning_done = True
1765
warning("Format %s for %s is deprecated - please use"
1766
" 'bzr upgrade --1.6.1-rich-root'"
1767
% (self._format, self.bzrdir.transport.base))
2288
1769
def _abort_write_group(self):
2289
self.revisions._index._key_dependencies.clear()
2290
1770
self._pack_collection._abort_write_group()
2292
def _get_source(self, to_format):
2293
if to_format.network_name() == self._format.network_name():
2294
return KnitPackStreamSource(self, to_format)
2295
return super(KnitPackRepository, self)._get_source(to_format)
1772
def _find_inconsistent_revision_parents(self):
1773
"""Find revisions with incorrectly cached parents.
1775
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1776
parents-in-revision).
1778
if not self.is_locked():
1779
raise errors.ObjectNotLocked(self)
1780
pb = ui.ui_factory.nested_progress_bar()
1783
revision_nodes = self._pack_collection.revision_index \
1784
.combined_index.iter_all_entries()
1785
index_positions = []
1786
# Get the cached index values for all revisions, and also the location
1787
# in each index of the revision text so we can perform linear IO.
1788
for index, key, value, refs in revision_nodes:
1789
pos, length = value[1:].split(' ')
1790
index_positions.append((index, int(pos), key[0],
1791
tuple(parent[0] for parent in refs[0])))
1792
pb.update("Reading revision index.", 0, 0)
1793
index_positions.sort()
1794
batch_count = len(index_positions) / 1000 + 1
1795
pb.update("Checking cached revision graph.", 0, batch_count)
1796
for offset in xrange(batch_count):
1797
pb.update("Checking cached revision graph.", offset)
1798
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
1801
rev_ids = [item[2] for item in to_query]
1802
revs = self.get_revisions(rev_ids)
1803
for revision, item in zip(revs, to_query):
1804
index_parents = item[3]
1805
rev_parents = tuple(revision.parent_ids)
1806
if index_parents != rev_parents:
1807
result.append((revision.revision_id, index_parents, rev_parents))
1812
@symbol_versioning.deprecated_method(symbol_versioning.one_one)
1813
def get_parents(self, revision_ids):
1814
"""See graph._StackedParentsProvider.get_parents."""
1815
parent_map = self.get_parent_map(revision_ids)
1816
return [parent_map.get(r, None) for r in revision_ids]
2297
1818
def _make_parents_provider(self):
2298
1819
return graph.CachingParentsProvider(self)
2300
1821
def _refresh_data(self):
2301
if not self.is_locked():
2303
self._pack_collection.reload_pack_names()
1822
if self._write_lock_count == 1 or (
1823
self.control_files._lock_count == 1 and
1824
self.control_files._lock_mode == 'r'):
1825
# forget what names there are
1826
self._pack_collection.reset()
1827
# XXX: Better to do an in-memory merge when acquiring a new lock -
1828
# factor out code from _save_pack_names.
1829
self._pack_collection.ensure_loaded()
2305
1831
def _start_write_group(self):
2306
1832
self._pack_collection._start_write_group()
2308
1834
def _commit_write_group(self):
2309
hint = self._pack_collection._commit_write_group()
2310
self.revisions._index._key_dependencies.clear()
2313
def suspend_write_group(self):
2314
# XXX check self._write_group is self.get_transaction()?
2315
tokens = self._pack_collection._suspend_write_group()
2316
self.revisions._index._key_dependencies.clear()
2317
self._write_group = None
2320
def _resume_write_group(self, tokens):
2321
self._start_write_group()
2323
self._pack_collection._resume_write_group(tokens)
2324
except errors.UnresumableWriteGroup:
2325
self._abort_write_group()
2327
for pack in self._pack_collection._resumed_packs:
2328
self.revisions._index.scan_unvalidated_index(pack.revision_index)
1835
return self._pack_collection._commit_write_group()
2330
1837
def get_transaction(self):
2331
1838
if self._write_lock_count:
2340
1847
return self._write_lock_count
2342
1849
def lock_write(self, token=None):
2343
locked = self.is_locked()
2344
if not self._write_lock_count and locked:
1850
if not self._write_lock_count and self.is_locked():
2345
1851
raise errors.ReadOnlyError(self)
2346
1852
self._write_lock_count += 1
2347
1853
if self._write_lock_count == 1:
2348
1854
self._transaction = transactions.WriteTransaction()
2350
if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2351
note('%r was write locked again', self)
2352
self._prev_lock = 'w'
2353
1855
for repo in self._fallback_repositories:
2354
1856
# Writes don't affect fallback repos
2355
1857
repo.lock_read()
2356
self._refresh_data()
1858
self._refresh_data()
2358
1860
def lock_read(self):
2359
locked = self.is_locked()
2360
1861
if self._write_lock_count:
2361
1862
self._write_lock_count += 1
2363
1864
self.control_files.lock_read()
2365
if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2366
note('%r was read locked again', self)
2367
self._prev_lock = 'r'
2368
1865
for repo in self._fallback_repositories:
1866
# Writes don't affect fallback repos
2369
1867
repo.lock_read()
2370
self._refresh_data()
1868
self._refresh_data()
2372
1870
def leave_lock_in_place(self):
2373
1871
# not supported - raise an error
2413
1906
transaction = self._transaction
2414
1907
self._transaction = None
2415
1908
transaction.finish()
1909
for repo in self._fallback_repositories:
2417
1912
self.control_files.unlock()
2419
if not self.is_locked():
2420
1913
for repo in self._fallback_repositories:
2424
class KnitPackStreamSource(StreamSource):
2425
"""A StreamSource used to transfer data between same-format KnitPack repos.
2427
This source assumes:
2428
1) Same serialization format for all objects
2429
2) Same root information
2430
3) XML format inventories
2431
4) Atomic inserts (so we can stream inventory texts before text
2436
def __init__(self, from_repository, to_format):
2437
super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2438
self._text_keys = None
2439
self._text_fetch_order = 'unordered'
2441
def _get_filtered_inv_stream(self, revision_ids):
2442
from_repo = self.from_repository
2443
parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2444
parent_keys = [(p,) for p in parent_ids]
2445
find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2446
parent_text_keys = set(find_text_keys(
2447
from_repo._inventory_xml_lines_for_keys(parent_keys)))
2448
content_text_keys = set()
2449
knit = KnitVersionedFiles(None, None)
2450
factory = KnitPlainFactory()
2451
def find_text_keys_from_content(record):
2452
if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2453
raise ValueError("Unknown content storage kind for"
2454
" inventory text: %s" % (record.storage_kind,))
2455
# It's a knit record, it has a _raw_record field (even if it was
2456
# reconstituted from a network stream).
2457
raw_data = record._raw_record
2458
# read the entire thing
2459
revision_id = record.key[-1]
2460
content, _ = knit._parse_record(revision_id, raw_data)
2461
if record.storage_kind == 'knit-delta-gz':
2462
line_iterator = factory.get_linedelta_content(content)
2463
elif record.storage_kind == 'knit-ft-gz':
2464
line_iterator = factory.get_fulltext_content(content)
2465
content_text_keys.update(find_text_keys(
2466
[(line, revision_id) for line in line_iterator]))
2467
revision_keys = [(r,) for r in revision_ids]
2468
def _filtered_inv_stream():
2469
source_vf = from_repo.inventories
2470
stream = source_vf.get_record_stream(revision_keys,
2472
for record in stream:
2473
if record.storage_kind == 'absent':
2474
raise errors.NoSuchRevision(from_repo, record.key)
2475
find_text_keys_from_content(record)
2477
self._text_keys = content_text_keys - parent_text_keys
2478
return ('inventories', _filtered_inv_stream())
2480
def _get_text_stream(self):
2481
# Note: We know we don't have to handle adding root keys, because both
2482
# the source and target are the identical network name.
2483
text_stream = self.from_repository.texts.get_record_stream(
2484
self._text_keys, self._text_fetch_order, False)
2485
return ('texts', text_stream)
2487
def get_stream(self, search):
2488
revision_ids = search.get_keys()
2489
for stream_info in self._fetch_revision_texts(revision_ids):
2491
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2492
yield self._get_filtered_inv_stream(revision_ids)
2493
yield self._get_text_stream()
2497
1917
class RepositoryFormatPack(MetaDirRepositoryFormat):
2498
1918
"""Format logic for pack structured repositories.
2832
2274
def get_format_string(self):
2833
2275
"""See RepositoryFormat.get_format_string()."""
2834
return "Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n"
2276
return "Bazaar development format 2 (needs bzr.dev from before 1.8)\n"
2836
2278
def get_format_description(self):
2837
2279
"""See RepositoryFormat.get_format_description()."""
2838
return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2841
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2842
"""A repository with rich roots, no subtrees, stacking and btree indexes.
2844
1.6-rich-root with B+Tree indices.
2847
repository_class = KnitPackRepository
2848
_commit_builder_class = PackRootCommitBuilder
2849
rich_root_data = True
2850
supports_tree_reference = False # no subtrees
2851
supports_external_lookups = True
2852
# What index classes to use
2853
index_builder_class = BTreeBuilder
2854
index_class = BTreeGraphIndex
2857
def _serializer(self):
2858
return xml6.serializer_v6
2860
def _get_matching_bzrdir(self):
2861
return bzrdir.format_registry.make_bzrdir(
2864
def _ignore_setting_bzrdir(self, format):
2280
return ("Development repository format, currently the same as "
2281
"1.6.1 with B+Trees.\n")
2283
def check_conversion_target(self, target_format):
2867
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2869
def get_format_string(self):
2870
"""See RepositoryFormat.get_format_string()."""
2871
return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2873
def get_format_description(self):
2874
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2877
2287
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2878
2288
"""A subtrees development repository.
added
removed
bzrlib/repofmt/pack_repo.py
