154
132
texts/deltas (via (fileid, revisionid) tuples).
155
133
:param signature_index: A GraphIndex for determining what signatures are
156
134
present in the Pack and accessing the locations of their texts.
157
:param chk_index: A GraphIndex for accessing content by CHK, if the
160
136
self.revision_index = revision_index
161
137
self.inventory_index = inventory_index
162
138
self.text_index = text_index
163
139
self.signature_index = signature_index
164
self.chk_index = chk_index
166
141
def access_tuple(self):
167
142
"""Return a tuple (transport, name) for the pack content."""
168
143
return self.pack_transport, self.file_name()
170
def _check_references(self):
171
"""Make sure our external references are present.
173
Packs are allowed to have deltas whose base is not in the pack, but it
174
must be present somewhere in this collection. It is not allowed to
175
have deltas based on a fallback repository.
176
(See <https://bugs.launchpad.net/bzr/+bug/288751>)
179
for (index_name, external_refs, index) in [
181
self._get_external_refs(self.text_index),
182
self._pack_collection.text_index.combined_index),
184
self._get_external_refs(self.inventory_index),
185
self._pack_collection.inventory_index.combined_index),
187
missing = external_refs.difference(
188
k for (idx, k, v, r) in
189
index.iter_entries(external_refs))
191
missing_items[index_name] = sorted(list(missing))
193
from pprint import pformat
194
raise errors.BzrCheckError(
195
"Newly created pack file %r has delta references to "
196
"items not in its repository:\n%s"
197
% (self, pformat(missing_items)))
199
145
def file_name(self):
200
146
"""Get the file name for the pack on disk."""
201
147
return self.name + '.pack'
227
165
"""The text index is the name + .tix."""
228
166
return self.index_name('text', name)
230
def _replace_index_with_readonly(self, index_type):
231
unlimited_cache = False
232
if index_type == 'chk':
233
unlimited_cache = True
234
setattr(self, index_type + '_index',
235
self.index_class(self.index_transport,
236
self.index_name(index_type, self.name),
237
self.index_sizes[self.index_offset(index_type)],
238
unlimited_cache=unlimited_cache))
168
def _external_compression_parents_of_texts(self):
171
for node in self.text_index.iter_all_entries():
173
refs.update(node[3][1])
241
177
class ExistingPack(Pack):
242
178
"""An in memory proxy for an existing .pack and its disk indices."""
244
180
def __init__(self, pack_transport, name, revision_index, inventory_index,
245
text_index, signature_index, chk_index=None):
181
text_index, signature_index):
246
182
"""Create an ExistingPack object.
248
184
:param pack_transport: The transport where the pack file resides.
249
185
:param name: The name of the pack on disk in the pack_transport.
251
187
Pack.__init__(self, revision_index, inventory_index, text_index,
252
signature_index, chk_index)
254
190
self.pack_transport = pack_transport
255
191
if None in (revision_index, inventory_index, text_index,
263
199
return not self.__eq__(other)
265
201
def __repr__(self):
266
return "<%s.%s object at 0x%x, %s, %s" % (
267
self.__class__.__module__, self.__class__.__name__, id(self),
268
self.pack_transport, self.name)
271
class ResumedPack(ExistingPack):
273
def __init__(self, name, revision_index, inventory_index, text_index,
274
signature_index, upload_transport, pack_transport, index_transport,
275
pack_collection, chk_index=None):
276
"""Create a ResumedPack object."""
277
ExistingPack.__init__(self, pack_transport, name, revision_index,
278
inventory_index, text_index, signature_index,
280
self.upload_transport = upload_transport
281
self.index_transport = index_transport
282
self.index_sizes = [None, None, None, None]
284
('revision', revision_index),
285
('inventory', inventory_index),
286
('text', text_index),
287
('signature', signature_index),
289
if chk_index is not None:
290
indices.append(('chk', chk_index))
291
self.index_sizes.append(None)
292
for index_type, index in indices:
293
offset = self.index_offset(index_type)
294
self.index_sizes[offset] = index._size
295
self.index_class = pack_collection._index_class
296
self._pack_collection = pack_collection
297
self._state = 'resumed'
298
# XXX: perhaps check that the .pack file exists?
300
def access_tuple(self):
301
if self._state == 'finished':
302
return Pack.access_tuple(self)
303
elif self._state == 'resumed':
304
return self.upload_transport, self.file_name()
306
raise AssertionError(self._state)
309
self.upload_transport.delete(self.file_name())
310
indices = [self.revision_index, self.inventory_index, self.text_index,
311
self.signature_index]
312
if self.chk_index is not None:
313
indices.append(self.chk_index)
314
for index in indices:
315
index._transport.delete(index._name)
318
self._check_references()
319
index_types = ['revision', 'inventory', 'text', 'signature']
320
if self.chk_index is not None:
321
index_types.append('chk')
322
for index_type in index_types:
323
old_name = self.index_name(index_type, self.name)
324
new_name = '../indices/' + old_name
325
self.upload_transport.rename(old_name, new_name)
326
self._replace_index_with_readonly(index_type)
327
new_name = '../packs/' + self.file_name()
328
self.upload_transport.rename(self.file_name(), new_name)
329
self._state = 'finished'
331
def _get_external_refs(self, index):
332
"""Return compression parents for this index that are not present.
334
This returns any compression parents that are referenced by this index,
335
which are not contained *in* this index. They may be present elsewhere.
337
return index.external_references(1)
202
return "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (
203
id(self), self.pack_transport, self.name)
340
206
class NewPack(Pack):
341
207
"""An in memory proxy for a pack which is being created."""
343
def __init__(self, pack_collection, upload_suffix='', file_mode=None):
209
# A map of index 'type' to the file extension and position in the
211
index_definitions = {
212
'revision': ('.rix', 0),
213
'inventory': ('.iix', 1),
215
'signature': ('.six', 3),
218
def __init__(self, upload_transport, index_transport, pack_transport,
219
upload_suffix='', file_mode=None):
344
220
"""Create a NewPack instance.
346
:param pack_collection: A PackCollection into which this is being inserted.
222
:param upload_transport: A writable transport for the pack to be
223
incrementally uploaded to.
224
:param index_transport: A writable transport for the pack's indices to
225
be written to when the pack is finished.
226
:param pack_transport: A writable transport for the pack to be renamed
227
to when the upload is complete. This *must* be the same as
228
upload_transport.clone('../packs').
347
229
:param upload_suffix: An optional suffix to be given to any temporary
348
230
files created during the pack creation. e.g '.autopack'
349
:param file_mode: Unix permissions for newly created file.
231
:param file_mode: An optional file mode to create the new files with.
351
233
# The relative locations of the packs are constrained, but all are
352
234
# passed in because the caller has them, so as to avoid object churn.
353
index_builder_class = pack_collection._index_builder_class
354
if pack_collection.chk_index is not None:
355
chk_index = index_builder_class(reference_lists=0)
358
235
Pack.__init__(self,
359
236
# Revisions: parents list, no text compression.
360
index_builder_class(reference_lists=1),
237
InMemoryGraphIndex(reference_lists=1),
361
238
# Inventory: We want to map compression only, but currently the
362
239
# knit code hasn't been updated enough to understand that, so we
363
240
# have a regular 2-list index giving parents and compression
365
index_builder_class(reference_lists=2),
242
InMemoryGraphIndex(reference_lists=2),
366
243
# Texts: compression and per file graph, for all fileids - so two
367
244
# reference lists and two elements in the key tuple.
368
index_builder_class(reference_lists=2, key_elements=2),
245
InMemoryGraphIndex(reference_lists=2, key_elements=2),
369
246
# Signatures: Just blobs to store, no compression, no parents
371
index_builder_class(reference_lists=0),
372
# CHK based storage - just blobs, no compression or parents.
248
InMemoryGraphIndex(reference_lists=0),
375
self._pack_collection = pack_collection
376
# When we make readonly indices, we need this.
377
self.index_class = pack_collection._index_class
378
250
# where should the new pack be opened
379
self.upload_transport = pack_collection._upload_transport
251
self.upload_transport = upload_transport
380
252
# where are indices written out to
381
self.index_transport = pack_collection._index_transport
253
self.index_transport = index_transport
382
254
# where is the pack renamed to when it is finished?
383
self.pack_transport = pack_collection._pack_transport
255
self.pack_transport = pack_transport
384
256
# What file mode to upload the pack and indices with.
385
257
self._file_mode = file_mode
386
258
# tracks the content written to the .pack file.
387
self._hash = osutils.md5()
388
# a tuple with the length in bytes of the indices, once the pack
389
# is finalised. (rev, inv, text, sigs, chk_if_in_use)
259
self._hash = md5.new()
260
# a four-tuple with the length in bytes of the indices, once the pack
261
# is finalised. (rev, inv, text, sigs)
390
262
self.index_sizes = None
391
263
# How much data to cache when writing packs. Note that this is not
392
264
# synchronised with reads, because it's not in the transport layer, so
477
338
- stores the index size tuple for the pack in the index_sizes
480
self.finish_content()
482
self._check_references()
343
self._write_data('', flush=True)
344
self.name = self._hash.hexdigest()
484
346
# XXX: It'd be better to write them all to temporary names, then
485
347
# rename them all into place, so that the window when only some are
486
348
# visible is smaller. On the other hand none will be seen until
487
349
# they're in the names list.
488
350
self.index_sizes = [None, None, None, None]
489
self._write_index('revision', self.revision_index, 'revision', suspend)
490
self._write_index('inventory', self.inventory_index, 'inventory',
492
self._write_index('text', self.text_index, 'file texts', suspend)
351
self._write_index('revision', self.revision_index, 'revision')
352
self._write_index('inventory', self.inventory_index, 'inventory')
353
self._write_index('text', self.text_index, 'file texts')
493
354
self._write_index('signature', self.signature_index,
494
'revision signatures', suspend)
495
if self.chk_index is not None:
496
self.index_sizes.append(None)
497
self._write_index('chk', self.chk_index,
498
'content hash bytes', suspend)
355
'revision signatures')
499
356
self.write_stream.close()
500
357
# Note that this will clobber an existing pack with the same name,
501
358
# without checking for hash collisions. While this is undesirable this
567
431
such as 'revision index'.
569
433
A CombinedIndex provides an index on a single key space built up
570
from several on-disk indices. The AggregateIndex builds on this
434
from several on-disk indices. The AggregateIndex builds on this
571
435
to provide a knit access layer, and allows having up to one writable
572
436
index within the collection.
574
438
# XXX: Probably 'can be written to' could/should be separated from 'acts
575
439
# like a knit index' -- mbp 20071024
577
def __init__(self, reload_func=None, flush_func=None):
578
"""Create an AggregateIndex.
580
:param reload_func: A function to call if we find we are missing an
581
index. Should have the form reload_func() => True if the list of
582
active pack files has changed.
584
self._reload_func = reload_func
442
"""Create an AggregateIndex."""
585
443
self.index_to_pack = {}
586
self.combined_index = CombinedGraphIndex([], reload_func=reload_func)
587
self.data_access = _DirectPackAccess(self.index_to_pack,
588
reload_func=reload_func,
589
flush_func=flush_func)
444
self.combined_index = CombinedGraphIndex([])
445
self.data_access = _DirectPackAccess(self.index_to_pack)
446
self.add_callback = None
448
def replace_indices(self, index_to_pack, indices):
449
"""Replace the current mappings with fresh ones.
451
This should probably not be used eventually, rather incremental add and
452
removal of indices. It has been added during refactoring of existing
455
:param index_to_pack: A mapping from index objects to
456
(transport, name) tuples for the pack file data.
457
:param indices: A list of indices.
459
# refresh the revision pack map dict without replacing the instance.
460
self.index_to_pack.clear()
461
self.index_to_pack.update(index_to_pack)
462
# XXX: API break - clearly a 'replace' method would be good?
463
self.combined_index._indices[:] = indices
464
# the current add nodes callback for the current writable index if
590
466
self.add_callback = None
592
468
def add_index(self, index, pack):
751
590
def open_pack(self):
752
591
"""Open a pack for the pack we are creating."""
753
new_pack = self._pack_collection.pack_factory(self._pack_collection,
754
upload_suffix=self.suffix,
755
file_mode=self._pack_collection.repo.bzrdir._get_file_mode())
756
# We know that we will process all nodes in order, and don't need to
757
# query, so don't combine any indices spilled to disk until we are done
758
new_pack.revision_index.set_optimize(combine_backing_indices=False)
759
new_pack.inventory_index.set_optimize(combine_backing_indices=False)
760
new_pack.text_index.set_optimize(combine_backing_indices=False)
761
new_pack.signature_index.set_optimize(combine_backing_indices=False)
764
def _update_pack_order(self, entries, index_to_pack_map):
765
"""Determine how we want our packs to be ordered.
767
This changes the sort order of the self.packs list so that packs unused
768
by 'entries' will be at the end of the list, so that future requests
769
can avoid probing them. Used packs will be at the front of the
770
self.packs list, in the order of their first use in 'entries'.
772
:param entries: A list of (index, ...) tuples
773
:param index_to_pack_map: A mapping from index objects to pack objects.
777
for entry in entries:
779
if index not in seen_indexes:
780
packs.append(index_to_pack_map[index])
781
seen_indexes.add(index)
782
if len(packs) == len(self.packs):
783
if 'pack' in debug.debug_flags:
784
mutter('Not changing pack list, all packs used.')
786
seen_packs = set(packs)
787
for pack in self.packs:
788
if pack not in seen_packs:
791
if 'pack' in debug.debug_flags:
792
old_names = [p.access_tuple()[1] for p in self.packs]
793
new_names = [p.access_tuple()[1] for p in packs]
794
mutter('Reordering packs\nfrom: %s\n to: %s',
795
old_names, new_names)
592
return NewPack(self._pack_collection._upload_transport,
593
self._pack_collection._index_transport,
594
self._pack_collection._pack_transport, upload_suffix=self.suffix,
595
file_mode=self._pack_collection.repo.bzrdir._get_file_mode())
798
597
def _copy_revision_texts(self):
799
598
"""Copy revision data to the new pack."""
942
740
self._pack_collection.allocate(new_pack)
945
def _copy_chks(self, refs=None):
946
# XXX: Todo, recursive follow-pointers facility when fetching some
948
chk_index_map, chk_indices = self._pack_map_and_index_list(
950
chk_nodes = self._index_contents(chk_indices, refs)
952
# TODO: This isn't strictly tasteful as we are accessing some private
953
# variables (_serializer). Perhaps a better way would be to have
954
# Repository._deserialise_chk_node()
955
search_key_func = chk_map.search_key_registry.get(
956
self._pack_collection.repo._serializer.search_key_name)
957
def accumlate_refs(lines):
958
# XXX: move to a generic location
960
bytes = ''.join(lines)
961
node = chk_map._deserialise(bytes, ("unknown",), search_key_func)
962
new_refs.update(node.refs())
963
self._copy_nodes(chk_nodes, chk_index_map, self.new_pack._writer,
964
self.new_pack.chk_index, output_lines=accumlate_refs)
967
def _copy_nodes(self, nodes, index_map, writer, write_index,
969
"""Copy knit nodes between packs with no graph references.
971
:param output_lines: Output full texts of copied items.
743
def _copy_nodes(self, nodes, index_map, writer, write_index):
744
"""Copy knit nodes between packs with no graph references."""
973
745
pb = ui.ui_factory.nested_progress_bar()
975
747
return self._do_copy_nodes(nodes, index_map, writer,
976
write_index, pb, output_lines=output_lines)
980
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb,
752
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
982
753
# for record verification
983
754
knit = KnitVersionedFiles(None, None)
984
755
# plan a readv on each source pack:
1005
776
# linear scan up the pack
1006
777
pack_readv_requests.sort()
1008
pack_obj = index_map[index]
1009
transport, path = pack_obj.access_tuple()
1011
reader = pack.make_readv_reader(transport, path,
1012
[offset[0:2] for offset in pack_readv_requests])
1013
except errors.NoSuchFile:
1014
if self._reload_func is not None:
779
transport, path = index_map[index]
780
reader = pack.make_readv_reader(transport, path,
781
[offset[0:2] for offset in pack_readv_requests])
1017
782
for (names, read_func), (_1, _2, (key, eol_flag)) in \
1018
783
izip(reader.iter_records(), pack_readv_requests):
1019
784
raw_data = read_func(None)
1020
785
# check the header only
1021
if output_lines is not None:
1022
output_lines(knit._parse_record(key[-1], raw_data)[0])
1024
df, _ = knit._parse_record_header(key, raw_data)
786
df, _ = knit._parse_record_header(key, raw_data)
1026
788
pos, size = writer.add_bytes_record(raw_data, names)
1027
789
write_index.add_node(key, eol_flag + "%d %d" % (pos, size))
1028
790
pb.update("Copied record", record_index)
1355
1101
class RepositoryPackCollection(object):
1356
1102
"""Management of packs within a repository.
1358
1104
:ivar _names: map of {pack_name: (index_size,)}
1361
pack_factory = NewPack
1362
resumed_pack_factory = ResumedPack
1364
1107
def __init__(self, repo, transport, index_transport, upload_transport,
1365
pack_transport, index_builder_class, index_class,
1367
1109
"""Create a new RepositoryPackCollection.
1369
:param transport: Addresses the repository base directory
1111
:param transport: Addresses the repository base directory
1370
1112
(typically .bzr/repository/).
1371
1113
:param index_transport: Addresses the directory containing indices.
1372
1114
:param upload_transport: Addresses the directory into which packs are written
1373
1115
while they're being created.
1374
1116
:param pack_transport: Addresses the directory of existing complete packs.
1375
:param index_builder_class: The index builder class to use.
1376
:param index_class: The index class to use.
1377
:param use_chk_index: Whether to setup and manage a CHK index.
1379
# XXX: This should call self.reset()
1380
1118
self.repo = repo
1381
1119
self.transport = transport
1382
1120
self._index_transport = index_transport
1383
1121
self._upload_transport = upload_transport
1384
1122
self._pack_transport = pack_transport
1385
self._index_builder_class = index_builder_class
1386
self._index_class = index_class
1387
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3,
1123
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1389
1124
self.packs = []
1390
1125
# name:Pack mapping
1392
1126
self._packs_by_name = {}
1393
1127
# the previous pack-names content
1394
1128
self._packs_at_load = None
1395
1129
# when a pack is being created by this object, the state of that pack.
1396
1130
self._new_pack = None
1397
1131
# aggregated revision index data
1398
flush = self._flush_new_pack
1399
self.revision_index = AggregateIndex(self.reload_pack_names, flush)
1400
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1401
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1402
self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1403
all_indices = [self.revision_index, self.inventory_index,
1404
self.text_index, self.signature_index]
1406
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1407
all_indices.append(self.chk_index)
1409
# used to determine if we're using a chk_index elsewhere.
1410
self.chk_index = None
1411
# Tell all the CombinedGraphIndex objects about each other, so they can
1412
# share hints about which pack names to search first.
1413
all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1414
for combined_idx in all_combined:
1415
combined_idx.set_sibling_indices(
1416
set(all_combined).difference([combined_idx]))
1418
self._resumed_packs = []
1421
return '%s(%r)' % (self.__class__.__name__, self.repo)
1132
self.revision_index = AggregateIndex()
1133
self.inventory_index = AggregateIndex()
1134
self.text_index = AggregateIndex()
1135
self.signature_index = AggregateIndex()
1423
1137
def add_pack_to_memory(self, pack):
1424
1138
"""Make a Pack object available to the repository to satisfy queries.
1426
1140
:param pack: A Pack object.
1428
1142
if pack.name in self._packs_by_name:
1429
raise AssertionError(
1430
'pack %s already in _packs_by_name' % (pack.name,))
1143
raise AssertionError()
1431
1144
self.packs.append(pack)
1432
1145
self._packs_by_name[pack.name] = pack
1433
1146
self.revision_index.add_index(pack.revision_index, pack)
1434
1147
self.inventory_index.add_index(pack.inventory_index, pack)
1435
1148
self.text_index.add_index(pack.text_index, pack)
1436
1149
self.signature_index.add_index(pack.signature_index, pack)
1437
if self.chk_index is not None:
1438
self.chk_index.add_index(pack.chk_index, pack)
1440
1151
def all_packs(self):
1441
1152
"""Return a list of all the Pack objects this repository has.
1492
1199
# group their data with the relevant commit, and that may
1493
1200
# involve rewriting ancient history - which autopack tries to
1494
1201
# avoid. Alternatively we could not group the data but treat
1495
# each of these as having a single revision, and thus add
1202
# each of these as having a single revision, and thus add
1496
1203
# one revision for each to the total revision count, to get
1497
1204
# a matching distribution.
1499
1206
existing_packs.append((revision_count, pack))
1500
1207
pack_operations = self.plan_autopack_combinations(
1501
1208
existing_packs, pack_distribution)
1502
num_new_packs = len(pack_operations)
1503
num_old_packs = sum([len(po[1]) for po in pack_operations])
1504
num_revs_affected = sum([po[0] for po in pack_operations])
1505
mutter('Auto-packing repository %s, which has %d pack files, '
1506
'containing %d revisions. Packing %d files into %d affecting %d'
1507
' revisions', self, total_packs, total_revisions, num_old_packs,
1508
num_new_packs, num_revs_affected)
1509
result = self._execute_pack_operations(pack_operations,
1510
reload_func=self._restart_autopack)
1511
mutter('Auto-packing repository %s completed', self)
1209
self._execute_pack_operations(pack_operations)
1514
def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1212
def _execute_pack_operations(self, pack_operations, _packer_class=Packer):
1516
1213
"""Execute a series of pack operations.
1518
1215
:param pack_operations: A list of [revision_count, packs_to_combine].
1519
1216
:param _packer_class: The class of packer to use (default: Packer).
1520
:return: The new pack names.
1522
1219
for revision_count, packs in pack_operations:
1523
1220
# we may have no-ops from the setup logic
1524
1221
if len(packs) == 0:
1526
packer = _packer_class(self, packs, '.autopack',
1527
reload_func=reload_func)
1530
except errors.RetryWithNewPacks:
1531
# An exception is propagating out of this context, make sure
1532
# this packer has cleaned up. Packer() doesn't set its new_pack
1533
# state into the RepositoryPackCollection object, so we only
1534
# have access to it directly here.
1535
if packer.new_pack is not None:
1536
packer.new_pack.abort()
1223
_packer_class(self, packs, '.autopack').pack()
1538
1224
for pack in packs:
1539
1225
self._remove_pack_from_memory(pack)
1540
1226
# record the newly available packs and stop advertising the old
1542
to_be_obsoleted = []
1543
for _, packs in pack_operations:
1544
to_be_obsoleted.extend(packs)
1545
result = self._save_pack_names(clear_obsolete_packs=True,
1546
obsolete_packs=to_be_obsoleted)
1549
def _flush_new_pack(self):
1550
if self._new_pack is not None:
1551
self._new_pack.flush()
1228
self._save_pack_names(clear_obsolete_packs=True)
1229
# Move the old packs out of the way now they are no longer referenced.
1230
for revision_count, packs in pack_operations:
1231
self._obsolete_packs(packs)
1553
1233
def lock_names(self):
1554
1234
"""Acquire the mutex around the pack-names index.
1556
1236
This cannot be used in the middle of a read-only transaction on the
1559
1239
self.repo.control_files.lock_write()
1561
def _already_packed(self):
1562
"""Is the collection already packed?"""
1563
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1565
def pack(self, hint=None, clean_obsolete_packs=False):
1566
1242
"""Pack the pack collection totally."""
1567
1243
self.ensure_loaded()
1568
1244
total_packs = len(self._names)
1569
if self._already_packed():
1246
# This is arguably wrong because we might not be optimal, but for
1247
# now lets leave it in. (e.g. reconcile -> one pack. But not
1571
1250
total_revisions = self.revision_index.combined_index.key_count()
1572
1251
# XXX: the following may want to be a class, to pack with a given
1574
1253
mutter('Packing repository %s, which has %d pack files, '
1575
'containing %d revisions with hint %r.', self, total_packs,
1576
total_revisions, hint)
1254
'containing %d revisions into 1 packs.', self, total_packs,
1577
1256
# determine which packs need changing
1257
pack_distribution = [1]
1578
1258
pack_operations = [[0, []]]
1579
1259
for pack in self.all_packs():
1580
if hint is None or pack.name in hint:
1581
# Either no hint was provided (so we are packing everything),
1582
# or this pack was included in the hint.
1583
pack_operations[-1][0] += pack.get_revision_count()
1584
pack_operations[-1][1].append(pack)
1260
pack_operations[-1][0] += pack.get_revision_count()
1261
pack_operations[-1][1].append(pack)
1585
1262
self._execute_pack_operations(pack_operations, OptimisingPacker)
1587
if clean_obsolete_packs:
1588
self._clear_obsolete_packs()
1590
1264
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1591
1265
"""Plan a pack operation.
1679
1335
inv_index = self._make_index(name, '.iix')
1680
1336
txt_index = self._make_index(name, '.tix')
1681
1337
sig_index = self._make_index(name, '.six')
1682
if self.chk_index is not None:
1683
chk_index = self._make_index(name, '.cix', unlimited_cache=True)
1686
1338
result = ExistingPack(self._pack_transport, name, rev_index,
1687
inv_index, txt_index, sig_index, chk_index)
1339
inv_index, txt_index, sig_index)
1688
1340
self.add_pack_to_memory(result)
1691
def _resume_pack(self, name):
1692
"""Get a suspended Pack object by name.
1694
:param name: The name of the pack - e.g. '123456'
1695
:return: A Pack object.
1697
if not re.match('[a-f0-9]{32}', name):
1698
# Tokens should be md5sums of the suspended pack file, i.e. 32 hex
1700
raise errors.UnresumableWriteGroup(
1701
self.repo, [name], 'Malformed write group token')
1703
rev_index = self._make_index(name, '.rix', resume=True)
1704
inv_index = self._make_index(name, '.iix', resume=True)
1705
txt_index = self._make_index(name, '.tix', resume=True)
1706
sig_index = self._make_index(name, '.six', resume=True)
1707
if self.chk_index is not None:
1708
chk_index = self._make_index(name, '.cix', resume=True,
1709
unlimited_cache=True)
1712
result = self.resumed_pack_factory(name, rev_index, inv_index,
1713
txt_index, sig_index, self._upload_transport,
1714
self._pack_transport, self._index_transport, self,
1715
chk_index=chk_index)
1716
except errors.NoSuchFile, e:
1717
raise errors.UnresumableWriteGroup(self.repo, [name], str(e))
1718
self.add_pack_to_memory(result)
1719
self._resumed_packs.append(result)
1722
1343
def allocate(self, a_new_pack):
1723
1344
"""Allocate name in the list of packs.
1735
1356
def _iter_disk_pack_index(self):
1736
1357
"""Iterate over the contents of the pack-names index.
1738
1359
This is used when loading the list from disk, and before writing to
1739
1360
detect updates from others during our write operation.
1740
1361
:return: An iterator of the index contents.
1742
return self._index_class(self.transport, 'pack-names', None
1363
return GraphIndex(self.transport, 'pack-names', None
1743
1364
).iter_all_entries()
1745
def _make_index(self, name, suffix, resume=False, unlimited_cache=False):
1366
def _make_index(self, name, suffix):
1746
1367
size_offset = self._suffix_offsets[suffix]
1747
1368
index_name = name + suffix
1749
transport = self._upload_transport
1750
index_size = transport.stat(index_name).st_size
1752
transport = self._index_transport
1753
index_size = self._names[name][size_offset]
1754
return self._index_class(transport, index_name, index_size,
1755
unlimited_cache=unlimited_cache)
1369
index_size = self._names[name][size_offset]
1371
self._index_transport, index_name, index_size)
1757
1373
def _max_pack_count(self, total_revisions):
1758
1374
"""Return the maximum number of packs to use for total revisions.
1760
1376
:param total_revisions: The total number of revisions in the
1786
1402
:param return: None.
1788
1404
for pack in packs:
1790
pack.pack_transport.rename(pack.file_name(),
1791
'../obsolete_packs/' + pack.file_name())
1792
except (errors.PathError, errors.TransportError), e:
1793
# TODO: Should these be warnings or mutters?
1794
mutter("couldn't rename obsolete pack, skipping it:\n%s"
1405
pack.pack_transport.rename(pack.file_name(),
1406
'../obsolete_packs/' + pack.file_name())
1796
1407
# TODO: Probably needs to know all possible indices for this pack
1797
1408
# - or maybe list the directory and move all indices matching this
1798
1409
# name whether we recognize it or not?
1799
suffixes = ['.iix', '.six', '.tix', '.rix']
1800
if self.chk_index is not None:
1801
suffixes.append('.cix')
1802
for suffix in suffixes:
1804
self._index_transport.rename(pack.name + suffix,
1805
'../obsolete_packs/' + pack.name + suffix)
1806
except (errors.PathError, errors.TransportError), e:
1807
mutter("couldn't rename obsolete index, skipping it:\n%s"
1410
for suffix in ('.iix', '.six', '.tix', '.rix'):
1411
self._index_transport.rename(pack.name + suffix,
1412
'../obsolete_packs/' + pack.name + suffix)
1810
1414
def pack_distribution(self, total_revisions):
1811
1415
"""Generate a list of the number of revisions to put in each pack.
1830
1434
def _remove_pack_from_memory(self, pack):
1831
1435
"""Remove pack from the packs accessed by this repository.
1833
1437
Only affects memory state, until self._save_pack_names() is invoked.
1835
1439
self._names.pop(pack.name)
1836
1440
self._packs_by_name.pop(pack.name)
1837
1441
self._remove_pack_indices(pack)
1838
self.packs.remove(pack)
1840
def _remove_pack_indices(self, pack, ignore_missing=False):
1841
"""Remove the indices for pack from the aggregated indices.
1843
:param ignore_missing: Suppress KeyErrors from calling remove_index.
1845
for index_type in Pack.index_definitions.keys():
1846
attr_name = index_type + '_index'
1847
aggregate_index = getattr(self, attr_name)
1848
if aggregate_index is not None:
1849
pack_index = getattr(pack, attr_name)
1851
aggregate_index.remove_index(pack_index)
1443
def _remove_pack_indices(self, pack):
1444
"""Remove the indices for pack from the aggregated indices."""
1445
self.revision_index.remove_index(pack.revision_index, pack)
1446
self.inventory_index.remove_index(pack.inventory_index, pack)
1447
self.text_index.remove_index(pack.text_index, pack)
1448
self.signature_index.remove_index(pack.signature_index, pack)
1857
1450
def reset(self):
1858
1451
"""Clear all cached data."""
1859
1452
# cached revision data
1453
self.repo._revision_knit = None
1860
1454
self.revision_index.clear()
1861
1455
# cached signature data
1456
self.repo._signature_knit = None
1862
1457
self.signature_index.clear()
1863
1458
# cached file text data
1864
1459
self.text_index.clear()
1460
self.repo._text_knit = None
1865
1461
# cached inventory data
1866
1462
self.inventory_index.clear()
1868
if self.chk_index is not None:
1869
self.chk_index.clear()
1870
1463
# remove the open pack
1871
1464
self._new_pack = None
1872
1465
# information about packs.
1875
1468
self._packs_by_name = {}
1876
1469
self._packs_at_load = None
1471
def _make_index_map(self, index_suffix):
1472
"""Return information on existing indices.
1474
:param suffix: Index suffix added to pack name.
1476
:returns: (pack_map, indices) where indices is a list of GraphIndex
1477
objects, and pack_map is a mapping from those objects to the
1478
pack tuple they describe.
1480
# TODO: stop using this; it creates new indices unnecessarily.
1481
self.ensure_loaded()
1482
suffix_map = {'.rix': 'revision_index',
1483
'.six': 'signature_index',
1484
'.iix': 'inventory_index',
1485
'.tix': 'text_index',
1487
return self._packs_list_to_pack_map_and_index_list(self.all_packs(),
1488
suffix_map[index_suffix])
1490
def _packs_list_to_pack_map_and_index_list(self, packs, index_attribute):
1491
"""Convert a list of packs to an index pack map and index list.
1493
:param packs: The packs list to process.
1494
:param index_attribute: The attribute that the desired index is found
1496
:return: A tuple (map, list) where map contains the dict from
1497
index:pack_tuple, and lsit contains the indices in the same order
1503
index = getattr(pack, index_attribute)
1504
indices.append(index)
1505
pack_map[index] = (pack.pack_transport, pack.file_name())
1506
return pack_map, indices
1508
def _index_contents(self, pack_map, key_filter=None):
1509
"""Get an iterable of the index contents from a pack_map.
1511
:param pack_map: A map from indices to pack details.
1512
:param key_filter: An optional filter to limit the
1515
indices = [index for index in pack_map.iterkeys()]
1516
all_index = CombinedGraphIndex(indices)
1517
if key_filter is None:
1518
return all_index.iter_all_entries()
1520
return all_index.iter_entries(key_filter)
1878
1522
def _unlock_names(self):
1879
1523
"""Release the mutex around the pack-names index."""
1880
1524
self.repo.control_files.unlock()
1882
def _diff_pack_names(self):
1883
"""Read the pack names from disk, and compare it to the one in memory.
1885
:return: (disk_nodes, deleted_nodes, new_nodes)
1886
disk_nodes The final set of nodes that should be referenced
1887
deleted_nodes Nodes which have been removed from when we started
1888
new_nodes Nodes that are newly introduced
1890
# load the disk nodes across
1892
for index, key, value in self._iter_disk_pack_index():
1893
disk_nodes.add((key, value))
1894
orig_disk_nodes = set(disk_nodes)
1896
# do a two-way diff against our original content
1897
current_nodes = set()
1898
for name, sizes in self._names.iteritems():
1900
((name, ), ' '.join(str(size) for size in sizes)))
1902
# Packs no longer present in the repository, which were present when we
1903
# locked the repository
1904
deleted_nodes = self._packs_at_load - current_nodes
1905
# Packs which this process is adding
1906
new_nodes = current_nodes - self._packs_at_load
1908
# Update the disk_nodes set to include the ones we are adding, and
1909
# remove the ones which were removed by someone else
1910
disk_nodes.difference_update(deleted_nodes)
1911
disk_nodes.update(new_nodes)
1913
return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
1915
def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1916
"""Given the correct set of pack files, update our saved info.
1918
:return: (removed, added, modified)
1919
removed pack names removed from self._names
1920
added pack names added to self._names
1921
modified pack names that had changed value
1926
## self._packs_at_load = disk_nodes
1526
def _save_pack_names(self, clear_obsolete_packs=False):
1527
"""Save the list of packs.
1529
This will take out the mutex around the pack names list for the
1530
duration of the method call. If concurrent updates have been made, a
1531
three-way merge between the current list and the current in memory list
1534
:param clear_obsolete_packs: If True, clear out the contents of the
1535
obsolete_packs directory.
1539
builder = GraphIndexBuilder()
1540
# load the disk nodes across
1542
for index, key, value in self._iter_disk_pack_index():
1543
disk_nodes.add((key, value))
1544
# do a two-way diff against our original content
1545
current_nodes = set()
1546
for name, sizes in self._names.iteritems():
1548
((name, ), ' '.join(str(size) for size in sizes)))
1549
deleted_nodes = self._packs_at_load - current_nodes
1550
new_nodes = current_nodes - self._packs_at_load
1551
disk_nodes.difference_update(deleted_nodes)
1552
disk_nodes.update(new_nodes)
1553
# TODO: handle same-name, index-size-changes here -
1554
# e.g. use the value from disk, not ours, *unless* we're the one
1556
for key, value in disk_nodes:
1557
builder.add_node(key, value)
1558
self.transport.put_file('pack-names', builder.finish(),
1559
mode=self.repo.bzrdir._get_file_mode())
1560
# move the baseline forward
1561
self._packs_at_load = disk_nodes
1562
if clear_obsolete_packs:
1563
self._clear_obsolete_packs()
1565
self._unlock_names()
1566
# synchronise the memory packs list with what we just wrote:
1927
1567
new_names = dict(disk_nodes)
1928
1568
# drop no longer present nodes
1929
1569
for pack in self.all_packs():
1930
1570
if (pack.name,) not in new_names:
1931
removed.append(pack.name)
1932
1571
self._remove_pack_from_memory(pack)
1933
1572
# add new nodes/refresh existing ones
1934
1573
for key, value in disk_nodes:
1948
1587
self._remove_pack_from_memory(self.get_pack_by_name(name))
1949
1588
self._names[name] = sizes
1950
1589
self.get_pack_by_name(name)
1951
modified.append(name)
1954
1592
self._names[name] = sizes
1955
1593
self.get_pack_by_name(name)
1957
return removed, added, modified
1959
def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
1960
"""Save the list of packs.
1962
This will take out the mutex around the pack names list for the
1963
duration of the method call. If concurrent updates have been made, a
1964
three-way merge between the current list and the current in memory list
1967
:param clear_obsolete_packs: If True, clear out the contents of the
1968
obsolete_packs directory.
1969
:param obsolete_packs: Packs that are obsolete once the new pack-names
1970
file has been written.
1971
:return: A list of the names saved that were not previously on disk.
1973
already_obsolete = []
1976
builder = self._index_builder_class()
1977
(disk_nodes, deleted_nodes, new_nodes,
1978
orig_disk_nodes) = self._diff_pack_names()
1979
# TODO: handle same-name, index-size-changes here -
1980
# e.g. use the value from disk, not ours, *unless* we're the one
1982
for key, value in disk_nodes:
1983
builder.add_node(key, value)
1984
self.transport.put_file('pack-names', builder.finish(),
1985
mode=self.repo.bzrdir._get_file_mode())
1986
self._packs_at_load = disk_nodes
1987
if clear_obsolete_packs:
1990
to_preserve = set([o.name for o in obsolete_packs])
1991
already_obsolete = self._clear_obsolete_packs(to_preserve)
1993
self._unlock_names()
1994
# synchronise the memory packs list with what we just wrote:
1995
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1997
# TODO: We could add one more condition here. "if o.name not in
1998
# orig_disk_nodes and o != the new_pack we haven't written to
1999
# disk yet. However, the new pack object is not easily
2000
# accessible here (it would have to be passed through the
2001
# autopacking code, etc.)
2002
obsolete_packs = [o for o in obsolete_packs
2003
if o.name not in already_obsolete]
2004
self._obsolete_packs(obsolete_packs)
2005
return [new_node[0][0] for new_node in new_nodes]
2007
def reload_pack_names(self):
2008
"""Sync our pack listing with what is present in the repository.
2010
This should be called when we find out that something we thought was
2011
present is now missing. This happens when another process re-packs the
2014
:return: True if the in-memory list of packs has been altered at all.
2016
# The ensure_loaded call is to handle the case where the first call
2017
# made involving the collection was to reload_pack_names, where we
2018
# don't have a view of disk contents. Its a bit of a bandaid, and
2019
# causes two reads of pack-names, but its a rare corner case not struck
2020
# with regular push/pull etc.
2021
first_read = self.ensure_loaded()
2024
# out the new value.
2025
(disk_nodes, deleted_nodes, new_nodes,
2026
orig_disk_nodes) = self._diff_pack_names()
2027
# _packs_at_load is meant to be the explicit list of names in
2028
# 'pack-names' at then start. As such, it should not contain any
2029
# pending names that haven't been written out yet.
2030
self._packs_at_load = orig_disk_nodes
2032
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2033
if removed or added or modified:
2037
def _restart_autopack(self):
2038
"""Reload the pack names list, and restart the autopack code."""
2039
if not self.reload_pack_names():
2040
# Re-raise the original exception, because something went missing
2041
# and a restart didn't find it
2043
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2045
def _clear_obsolete_packs(self, preserve=None):
1595
def _clear_obsolete_packs(self):
2046
1596
"""Delete everything from the obsolete-packs directory.
2048
:return: A list of pack identifiers (the filename without '.pack') that
2049
were found in obsolete_packs.
2052
1598
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2053
if preserve is None:
2055
1599
for filename in obsolete_pack_transport.list_dir('.'):
2056
name, ext = osutils.splitext(filename)
2059
if name in preserve:
2062
1601
obsolete_pack_transport.delete(filename)
2063
1602
except (errors.PathError, errors.TransportError), e:
2064
warning("couldn't delete obsolete pack, skipping it:\n%s"
1603
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2068
1605
def _start_write_group(self):
2069
1606
# Do not permit preparation for writing if we're not in a 'write lock'.
2070
1607
if not self.repo.is_write_locked():
2071
1608
raise errors.NotWriteLocked(self)
2072
self._new_pack = self.pack_factory(self, upload_suffix='.pack',
1609
self._new_pack = NewPack(self._upload_transport, self._index_transport,
1610
self._pack_transport, upload_suffix='.pack',
2073
1611
file_mode=self.repo.bzrdir._get_file_mode())
2074
1612
# allow writing: queue writes to a new index
2075
1613
self.revision_index.add_writable_index(self._new_pack.revision_index,
2096
1628
# FIXME: just drop the transient index.
2097
1629
# forget what names there are
2098
1630
if self._new_pack is not None:
2099
operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2100
operation.add_cleanup(setattr, self, '_new_pack', None)
2101
# If we aborted while in the middle of finishing the write
2102
# group, _remove_pack_indices could fail because the indexes are
2103
# already gone. But they're not there we shouldn't fail in this
2104
# case, so we pass ignore_missing=True.
2105
operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2106
ignore_missing=True)
2107
operation.run_simple()
2108
for resumed_pack in self._resumed_packs:
2109
operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2110
# See comment in previous finally block.
2111
operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2112
ignore_missing=True)
2113
operation.run_simple()
2114
del self._resumed_packs[:]
2116
def _remove_resumed_pack_indices(self):
2117
for resumed_pack in self._resumed_packs:
2118
self._remove_pack_indices(resumed_pack)
2119
del self._resumed_packs[:]
2121
def _check_new_inventories(self):
2122
"""Detect missing inventories in this write group.
2124
:returns: list of strs, summarising any problems found. If the list is
2125
empty no problems were found.
2127
# The base implementation does no checks. GCRepositoryPackCollection
1631
self._new_pack.abort()
1632
self._remove_pack_indices(self._new_pack)
1633
self._new_pack = None
1634
self.repo._text_knit = None
2131
1636
def _commit_write_group(self):
2133
for prefix, versioned_file in (
2134
('revisions', self.repo.revisions),
2135
('inventories', self.repo.inventories),
2136
('texts', self.repo.texts),
2137
('signatures', self.repo.signatures),
2139
missing = versioned_file.get_missing_compression_parent_keys()
2140
all_missing.update([(prefix,) + key for key in missing])
2142
raise errors.BzrCheckError(
2143
"Repository %s has missing compression parent(s) %r "
2144
% (self.repo, sorted(all_missing)))
2145
problems = self._check_new_inventories()
2147
problems_summary = '\n'.join(problems)
2148
raise errors.BzrCheckError(
2149
"Cannot add revision(s) to repository: " + problems_summary)
2150
1637
self._remove_pack_indices(self._new_pack)
2151
any_new_content = False
2152
1638
if self._new_pack.data_inserted():
2153
1639
# get all the data to disk and read to use
2154
1640
self._new_pack.finish()
2155
1641
self.allocate(self._new_pack)
2156
1642
self._new_pack = None
2157
any_new_content = True
2159
self._new_pack.abort()
2160
self._new_pack = None
2161
for resumed_pack in self._resumed_packs:
2162
# XXX: this is a pretty ugly way to turn the resumed pack into a
2163
# properly committed pack.
2164
self._names[resumed_pack.name] = None
2165
self._remove_pack_from_memory(resumed_pack)
2166
resumed_pack.finish()
2167
self.allocate(resumed_pack)
2168
any_new_content = True
2169
del self._resumed_packs[:]
2171
result = self.autopack()
1643
if not self.autopack():
2173
1644
# when autopack takes no steps, the names list is still
2175
return self._save_pack_names()
2179
def _suspend_write_group(self):
2180
tokens = [pack.name for pack in self._resumed_packs]
2181
self._remove_pack_indices(self._new_pack)
2182
if self._new_pack.data_inserted():
2183
# get all the data to disk and read to use
2184
self._new_pack.finish(suspend=True)
2185
tokens.append(self._new_pack.name)
2186
self._new_pack = None
1646
self._save_pack_names()
2188
1648
self._new_pack.abort()
2189
1649
self._new_pack = None
2190
self._remove_resumed_pack_indices()
2193
def _resume_write_group(self, tokens):
2194
for token in tokens:
2195
self._resume_pack(token)
1650
self.repo._text_knit = None
2198
1653
class KnitPackRepository(KnitRepository):
2199
1654
"""Repository with knit objects stored inside pack containers.
2201
1656
The layering for a KnitPackRepository is:
2203
1658
Graph | HPSS | Repository public layer |
2281
1719
self._reconcile_does_inventory_gc = True
2282
1720
self._reconcile_fixes_text_parents = True
2283
1721
self._reconcile_backsup_inventory = False
1722
self._fetch_order = 'unordered'
2285
def _warn_if_deprecated(self, branch=None):
1724
def _warn_if_deprecated(self):
2286
1725
# This class isn't deprecated, but one sub-format is
2287
1726
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2288
super(KnitPackRepository, self)._warn_if_deprecated(branch)
1727
from bzrlib import repository
1728
if repository._deprecation_warning_done:
1730
repository._deprecation_warning_done = True
1731
warning("Format %s for %s is deprecated - please use"
1732
" 'bzr upgrade --1.6.1-rich-root'"
1733
% (self._format, self.bzrdir.transport.base))
2290
1735
def _abort_write_group(self):
2291
self.revisions._index._key_dependencies.clear()
2292
1736
self._pack_collection._abort_write_group()
2294
def _get_source(self, to_format):
2295
if to_format.network_name() == self._format.network_name():
2296
return KnitPackStreamSource(self, to_format)
2297
return super(KnitPackRepository, self)._get_source(to_format)
1738
def _find_inconsistent_revision_parents(self):
1739
"""Find revisions with incorrectly cached parents.
1741
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1742
parents-in-revision).
1744
if not self.is_locked():
1745
raise errors.ObjectNotLocked(self)
1746
pb = ui.ui_factory.nested_progress_bar()
1749
revision_nodes = self._pack_collection.revision_index \
1750
.combined_index.iter_all_entries()
1751
index_positions = []
1752
# Get the cached index values for all revisions, and also the location
1753
# in each index of the revision text so we can perform linear IO.
1754
for index, key, value, refs in revision_nodes:
1755
pos, length = value[1:].split(' ')
1756
index_positions.append((index, int(pos), key[0],
1757
tuple(parent[0] for parent in refs[0])))
1758
pb.update("Reading revision index.", 0, 0)
1759
index_positions.sort()
1760
batch_count = len(index_positions) / 1000 + 1
1761
pb.update("Checking cached revision graph.", 0, batch_count)
1762
for offset in xrange(batch_count):
1763
pb.update("Checking cached revision graph.", offset)
1764
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
1767
rev_ids = [item[2] for item in to_query]
1768
revs = self.get_revisions(rev_ids)
1769
for revision, item in zip(revs, to_query):
1770
index_parents = item[3]
1771
rev_parents = tuple(revision.parent_ids)
1772
if index_parents != rev_parents:
1773
result.append((revision.revision_id, index_parents, rev_parents))
1778
@symbol_versioning.deprecated_method(symbol_versioning.one_one)
1779
def get_parents(self, revision_ids):
1780
"""See graph._StackedParentsProvider.get_parents."""
1781
parent_map = self.get_parent_map(revision_ids)
1782
return [parent_map.get(r, None) for r in revision_ids]
2299
1784
def _make_parents_provider(self):
2300
1785
return graph.CachingParentsProvider(self)
2302
1787
def _refresh_data(self):
2303
if not self.is_locked():
2305
self._pack_collection.reload_pack_names()
1788
if self._write_lock_count == 1 or (
1789
self.control_files._lock_count == 1 and
1790
self.control_files._lock_mode == 'r'):
1791
# forget what names there are
1792
self._pack_collection.reset()
1793
# XXX: Better to do an in-memory merge when acquiring a new lock -
1794
# factor out code from _save_pack_names.
1795
self._pack_collection.ensure_loaded()
2307
1797
def _start_write_group(self):
2308
1798
self._pack_collection._start_write_group()
2310
1800
def _commit_write_group(self):
2311
hint = self._pack_collection._commit_write_group()
2312
self.revisions._index._key_dependencies.clear()
2315
def suspend_write_group(self):
2316
# XXX check self._write_group is self.get_transaction()?
2317
tokens = self._pack_collection._suspend_write_group()
2318
self.revisions._index._key_dependencies.clear()
2319
self._write_group = None
2322
def _resume_write_group(self, tokens):
2323
self._start_write_group()
2325
self._pack_collection._resume_write_group(tokens)
2326
except errors.UnresumableWriteGroup:
2327
self._abort_write_group()
2329
for pack in self._pack_collection._resumed_packs:
2330
self.revisions._index.scan_unvalidated_index(pack.revision_index)
1801
return self._pack_collection._commit_write_group()
2332
1803
def get_transaction(self):
2333
1804
if self._write_lock_count:
2342
1813
return self._write_lock_count
2344
1815
def lock_write(self, token=None):
2345
"""Lock the repository for writes.
2347
:return: A bzrlib.repository.RepositoryWriteLockResult.
2349
locked = self.is_locked()
2350
if not self._write_lock_count and locked:
1816
if not self._write_lock_count and self.is_locked():
2351
1817
raise errors.ReadOnlyError(self)
2352
1818
self._write_lock_count += 1
2353
1819
if self._write_lock_count == 1:
2354
1820
self._transaction = transactions.WriteTransaction()
2356
if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2357
note('%r was write locked again', self)
2358
self._prev_lock = 'w'
2359
1821
for repo in self._fallback_repositories:
2360
1822
# Writes don't affect fallback repos
2361
1823
repo.lock_read()
2362
self._refresh_data()
2363
return RepositoryWriteLockResult(self.unlock, None)
1824
self._refresh_data()
2365
1826
def lock_read(self):
2366
"""Lock the repository for reads.
2368
:return: A bzrlib.lock.LogicalLockResult.
2370
locked = self.is_locked()
2371
1827
if self._write_lock_count:
2372
1828
self._write_lock_count += 1
2374
1830
self.control_files.lock_read()
2376
if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2377
note('%r was read locked again', self)
2378
self._prev_lock = 'r'
2379
1831
for repo in self._fallback_repositories:
1832
# Writes don't affect fallback repos
2380
1833
repo.lock_read()
2381
self._refresh_data()
2382
return LogicalLockResult(self.unlock)
1834
self._refresh_data()
2384
1836
def leave_lock_in_place(self):
2385
1837
# not supported - raise an error
2425
1872
transaction = self._transaction
2426
1873
self._transaction = None
2427
1874
transaction.finish()
1875
for repo in self._fallback_repositories:
2429
1878
self.control_files.unlock()
2431
if not self.is_locked():
2432
1879
for repo in self._fallback_repositories:
2436
class KnitPackStreamSource(StreamSource):
2437
"""A StreamSource used to transfer data between same-format KnitPack repos.
2439
This source assumes:
2440
1) Same serialization format for all objects
2441
2) Same root information
2442
3) XML format inventories
2443
4) Atomic inserts (so we can stream inventory texts before text
2448
def __init__(self, from_repository, to_format):
2449
super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2450
self._text_keys = None
2451
self._text_fetch_order = 'unordered'
2453
def _get_filtered_inv_stream(self, revision_ids):
2454
from_repo = self.from_repository
2455
parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2456
parent_keys = [(p,) for p in parent_ids]
2457
find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2458
parent_text_keys = set(find_text_keys(
2459
from_repo._inventory_xml_lines_for_keys(parent_keys)))
2460
content_text_keys = set()
2461
knit = KnitVersionedFiles(None, None)
2462
factory = KnitPlainFactory()
2463
def find_text_keys_from_content(record):
2464
if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2465
raise ValueError("Unknown content storage kind for"
2466
" inventory text: %s" % (record.storage_kind,))
2467
# It's a knit record, it has a _raw_record field (even if it was
2468
# reconstituted from a network stream).
2469
raw_data = record._raw_record
2470
# read the entire thing
2471
revision_id = record.key[-1]
2472
content, _ = knit._parse_record(revision_id, raw_data)
2473
if record.storage_kind == 'knit-delta-gz':
2474
line_iterator = factory.get_linedelta_content(content)
2475
elif record.storage_kind == 'knit-ft-gz':
2476
line_iterator = factory.get_fulltext_content(content)
2477
content_text_keys.update(find_text_keys(
2478
[(line, revision_id) for line in line_iterator]))
2479
revision_keys = [(r,) for r in revision_ids]
2480
def _filtered_inv_stream():
2481
source_vf = from_repo.inventories
2482
stream = source_vf.get_record_stream(revision_keys,
2484
for record in stream:
2485
if record.storage_kind == 'absent':
2486
raise errors.NoSuchRevision(from_repo, record.key)
2487
find_text_keys_from_content(record)
2489
self._text_keys = content_text_keys - parent_text_keys
2490
return ('inventories', _filtered_inv_stream())
2492
def _get_text_stream(self):
2493
# Note: We know we don't have to handle adding root keys, because both
2494
# the source and target are the identical network name.
2495
text_stream = self.from_repository.texts.get_record_stream(
2496
self._text_keys, self._text_fetch_order, False)
2497
return ('texts', text_stream)
2499
def get_stream(self, search):
2500
revision_ids = search.get_keys()
2501
for stream_info in self._fetch_revision_texts(revision_ids):
2503
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2504
yield self._get_filtered_inv_stream(revision_ids)
2505
yield self._get_text_stream()
2509
1883
class RepositoryFormatPack(MetaDirRepositoryFormat):
2510
1884
"""Format logic for pack structured repositories.
2847
2199
def get_format_string(self):
2848
2200
"""See RepositoryFormat.get_format_string()."""
2849
return "Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n"
2201
return "Bazaar development format 1 (needs bzr.dev from before 1.6)\n"
2851
2203
def get_format_description(self):
2852
2204
"""See RepositoryFormat.get_format_description()."""
2853
return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2856
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2857
"""A repository with rich roots, no subtrees, stacking and btree indexes.
2859
1.6-rich-root with B+Tree indices.
2862
repository_class = KnitPackRepository
2863
_commit_builder_class = PackRootCommitBuilder
2864
rich_root_data = True
2865
supports_tree_reference = False # no subtrees
2866
supports_external_lookups = True
2867
# What index classes to use
2868
index_builder_class = BTreeBuilder
2869
index_class = BTreeGraphIndex
2872
def _serializer(self):
2873
return xml6.serializer_v6
2875
def _get_matching_bzrdir(self):
2876
return bzrdir.format_registry.make_bzrdir(
2879
def _ignore_setting_bzrdir(self, format):
2205
return ("Development repository format, currently the same as "
2206
"pack-0.92 with external reference support.\n")
2208
def check_conversion_target(self, target_format):
2882
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2884
def get_format_string(self):
2885
"""See RepositoryFormat.get_format_string()."""
2886
return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2888
def get_format_description(self):
2889
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2892
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2212
class RepositoryFormatPackDevelopment1Subtree(RepositoryFormatPack):
2893
2213
"""A subtrees development repository.
2895
This format should be retained until the second release after bzr 1.7.
2897
1.6.1-subtree[as it might have been] with B+Tree indices.
2899
This is [now] retained until we have a CHK based subtree format in
2215
This format should be retained until the second release after bzr 1.5.
2217
Supports external lookups, which results in non-truncated ghosts after
2218
reconcile compared to pack-0.92 formats.
2903
2221
repository_class = KnitPackRepository
2904
2222
_commit_builder_class = PackRootCommitBuilder
2905
2223
rich_root_data = True
2907
2224
supports_tree_reference = True
2225
_serializer = xml7.serializer_v7
2908
2226
supports_external_lookups = True
2909
# What index classes to use
2910
index_builder_class = BTreeBuilder
2911
index_class = BTreeGraphIndex
2914
def _serializer(self):
2915
return xml7.serializer_v7
2917
2228
def _get_matching_bzrdir(self):
2918
2229
return bzrdir.format_registry.make_bzrdir(
2919
'development-subtree')
2230
'development1-subtree')
2921
2232
def _ignore_setting_bzrdir(self, format):
2924
2235
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2237
def check_conversion_target(self, target_format):
2238
if not target_format.rich_root_data:
2239
raise errors.BadConversionTarget(
2240
'Does not support rich root data.', target_format)
2241
if not getattr(target_format, 'supports_tree_reference', False):
2242
raise errors.BadConversionTarget(
2243
'Does not support nested trees', target_format)
2926
2245
def get_format_string(self):
2927
2246
"""See RepositoryFormat.get_format_string()."""
2928
return ("Bazaar development format 2 with subtree support "
2929
"(needs bzr.dev from before 1.8)\n")
2247
return ("Bazaar development format 1 with subtree support "
2248
"(needs bzr.dev from before 1.6)\n")
2931
2250
def get_format_description(self):
2932
2251
"""See RepositoryFormat.get_format_description()."""
2933
2252
return ("Development repository format, currently the same as "
2934
"1.6.1-subtree with B+Tree indices.\n")
2253
"pack-0.92-subtree with external reference support.\n")
added
removed
bzrlib/repofmt/pack_repo.py
