262
207
return not self.__eq__(other)
264
209
def __repr__(self):
265
return "<%s.%s object at 0x%x, %s, %s" % (
266
self.__class__.__module__, self.__class__.__name__, id(self),
267
self.pack_transport, self.name)
270
class ResumedPack(ExistingPack):
272
def __init__(self, name, revision_index, inventory_index, text_index,
273
signature_index, upload_transport, pack_transport, index_transport,
274
pack_collection, chk_index=None):
275
"""Create a ResumedPack object."""
276
ExistingPack.__init__(self, pack_transport, name, revision_index,
277
inventory_index, text_index, signature_index,
279
self.upload_transport = upload_transport
280
self.index_transport = index_transport
281
self.index_sizes = [None, None, None, None]
283
('revision', revision_index),
284
('inventory', inventory_index),
285
('text', text_index),
286
('signature', signature_index),
288
if chk_index is not None:
289
indices.append(('chk', chk_index))
290
self.index_sizes.append(None)
291
for index_type, index in indices:
292
offset = self.index_offset(index_type)
293
self.index_sizes[offset] = index._size
294
self.index_class = pack_collection._index_class
295
self._pack_collection = pack_collection
296
self._state = 'resumed'
297
# XXX: perhaps check that the .pack file exists?
299
def access_tuple(self):
300
if self._state == 'finished':
301
return Pack.access_tuple(self)
302
elif self._state == 'resumed':
303
return self.upload_transport, self.file_name()
305
raise AssertionError(self._state)
308
self.upload_transport.delete(self.file_name())
309
indices = [self.revision_index, self.inventory_index, self.text_index,
310
self.signature_index]
311
if self.chk_index is not None:
312
indices.append(self.chk_index)
313
for index in indices:
314
index._transport.delete(index._name)
317
self._check_references()
318
index_types = ['revision', 'inventory', 'text', 'signature']
319
if self.chk_index is not None:
320
index_types.append('chk')
321
for index_type in index_types:
322
old_name = self.index_name(index_type, self.name)
323
new_name = '../indices/' + old_name
324
self.upload_transport.rename(old_name, new_name)
325
self._replace_index_with_readonly(index_type)
326
new_name = '../packs/' + self.file_name()
327
self.upload_transport.rename(self.file_name(), new_name)
328
self._state = 'finished'
330
def _get_external_refs(self, index):
331
"""Return compression parents for this index that are not present.
333
This returns any compression parents that are referenced by this index,
334
which are not contained *in* this index. They may be present elsewhere.
336
return index.external_references(1)
210
return "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (
211
id(self), self.transport, self.name)
339
214
class NewPack(Pack):
340
215
"""An in memory proxy for a pack which is being created."""
342
def __init__(self, pack_collection, upload_suffix='', file_mode=None):
217
# A map of index 'type' to the file extension and position in the
219
index_definitions = {
220
'revision': ('.rix', 0),
221
'inventory': ('.iix', 1),
223
'signature': ('.six', 3),
226
def __init__(self, upload_transport, index_transport, pack_transport,
227
upload_suffix='', file_mode=None):
343
228
"""Create a NewPack instance.
345
: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').
346
237
:param upload_suffix: An optional suffix to be given to any temporary
347
238
files created during the pack creation. e.g '.autopack'
348
:param file_mode: Unix permissions for newly created file.
239
:param file_mode: An optional file mode to create the new files with.
350
241
# The relative locations of the packs are constrained, but all are
351
242
# passed in because the caller has them, so as to avoid object churn.
352
index_builder_class = pack_collection._index_builder_class
353
if pack_collection.chk_index is not None:
354
chk_index = index_builder_class(reference_lists=0)
357
243
Pack.__init__(self,
358
244
# Revisions: parents list, no text compression.
359
index_builder_class(reference_lists=1),
245
InMemoryGraphIndex(reference_lists=1),
360
246
# Inventory: We want to map compression only, but currently the
361
247
# knit code hasn't been updated enough to understand that, so we
362
248
# have a regular 2-list index giving parents and compression
364
index_builder_class(reference_lists=2),
250
InMemoryGraphIndex(reference_lists=2),
365
251
# Texts: compression and per file graph, for all fileids - so two
366
252
# reference lists and two elements in the key tuple.
367
index_builder_class(reference_lists=2, key_elements=2),
253
InMemoryGraphIndex(reference_lists=2, key_elements=2),
368
254
# Signatures: Just blobs to store, no compression, no parents
370
index_builder_class(reference_lists=0),
371
# CHK based storage - just blobs, no compression or parents.
256
InMemoryGraphIndex(reference_lists=0),
374
self._pack_collection = pack_collection
375
# When we make readonly indices, we need this.
376
self.index_class = pack_collection._index_class
377
258
# where should the new pack be opened
378
self.upload_transport = pack_collection._upload_transport
259
self.upload_transport = upload_transport
379
260
# where are indices written out to
380
self.index_transport = pack_collection._index_transport
261
self.index_transport = index_transport
381
262
# where is the pack renamed to when it is finished?
382
self.pack_transport = pack_collection._pack_transport
263
self.pack_transport = pack_transport
383
264
# What file mode to upload the pack and indices with.
384
265
self._file_mode = file_mode
385
266
# tracks the content written to the .pack file.
386
self._hash = osutils.md5()
387
# a tuple with the length in bytes of the indices, once the pack
388
# is finalised. (rev, inv, text, sigs, chk_if_in_use)
267
self._hash = md5.new()
268
# a four-tuple with the length in bytes of the indices, once the pack
269
# is finalised. (rev, inv, text, sigs)
389
270
self.index_sizes = None
390
271
# How much data to cache when writing packs. Note that this is not
391
272
# synchronised with reads, because it's not in the transport layer, so
594
477
Future searches on the aggregate index will seach this new index
595
478
before all previously inserted indices.
597
480
:param index: An Index for the pack.
598
481
:param pack: A Pack instance.
600
483
# expose it to the index map
601
484
self.index_to_pack[index] = pack.access_tuple()
602
485
# put it at the front of the linear index list
603
self.combined_index.insert_index(0, index, pack.name)
486
self.combined_index.insert_index(0, index)
605
488
def add_writable_index(self, index, pack):
606
489
"""Add an index which is able to have data added to it.
608
491
There can be at most one writable index at any time. Any
609
492
modifications made to the knit are put into this index.
611
494
:param index: An index from the pack parameter.
612
495
:param pack: A Pack instance.
614
if self.add_callback is not None:
615
raise AssertionError(
616
"%s already has a writable index through %s" % \
617
(self, self.add_callback))
497
assert self.add_callback is None, \
498
"%s already has a writable index through %s" % \
499
(self, self.add_callback)
618
500
# allow writing: queue writes to a new index
619
501
self.add_index(index, pack)
620
502
# Updates the index to packs mapping as a side effect,
621
self.data_access.set_writer(pack._writer, index, pack.access_tuple())
503
self.knit_access.set_writer(pack._writer, index, pack.access_tuple())
622
504
self.add_callback = index.add_nodes
625
507
"""Reset all the aggregate data to nothing."""
626
self.data_access.set_writer(None, None, (None, None))
508
self.knit_access.set_writer(None, None, (None, None))
627
509
self.index_to_pack.clear()
628
510
del self.combined_index._indices[:]
629
del self.combined_index._index_names[:]
630
511
self.add_callback = None
632
def remove_index(self, index):
513
def remove_index(self, index, pack):
633
514
"""Remove index from the indices used to answer queries.
635
516
:param index: An index from the pack parameter.
517
:param pack: A Pack instance.
637
519
del self.index_to_pack[index]
638
pos = self.combined_index._indices.index(index)
639
del self.combined_index._indices[pos]
640
del self.combined_index._index_names[pos]
520
self.combined_index._indices.remove(index)
641
521
if (self.add_callback is not None and
642
522
getattr(index, 'add_nodes', None) == self.add_callback):
643
523
self.add_callback = None
644
self.data_access.set_writer(None, None, (None, None))
524
self.knit_access.set_writer(None, None, (None, None))
647
527
class Packer(object):
648
528
"""Create a pack from packs."""
650
def __init__(self, pack_collection, packs, suffix, revision_ids=None,
530
def __init__(self, pack_collection, packs, suffix, revision_ids=None):
652
531
"""Create a Packer.
654
533
:param pack_collection: A RepositoryPackCollection object where the
1354
1112
class RepositoryPackCollection(object):
1355
"""Management of packs within a repository.
1357
:ivar _names: map of {pack_name: (index_size,)}
1360
pack_factory = NewPack
1361
resumed_pack_factory = ResumedPack
1113
"""Management of packs within a repository."""
1363
1115
def __init__(self, repo, transport, index_transport, upload_transport,
1364
pack_transport, index_builder_class, index_class,
1366
1117
"""Create a new RepositoryPackCollection.
1368
:param transport: Addresses the repository base directory
1119
:param transport: Addresses the repository base directory
1369
1120
(typically .bzr/repository/).
1370
1121
:param index_transport: Addresses the directory containing indices.
1371
1122
:param upload_transport: Addresses the directory into which packs are written
1372
1123
while they're being created.
1373
1124
:param pack_transport: Addresses the directory of existing complete packs.
1374
:param index_builder_class: The index builder class to use.
1375
:param index_class: The index class to use.
1376
:param use_chk_index: Whether to setup and manage a CHK index.
1378
# XXX: This should call self.reset()
1379
1126
self.repo = repo
1380
1127
self.transport = transport
1381
1128
self._index_transport = index_transport
1382
1129
self._upload_transport = upload_transport
1383
1130
self._pack_transport = pack_transport
1384
self._index_builder_class = index_builder_class
1385
self._index_class = index_class
1386
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3,
1131
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1388
1132
self.packs = []
1389
1133
# name:Pack mapping
1391
1134
self._packs_by_name = {}
1392
1135
# the previous pack-names content
1393
1136
self._packs_at_load = None
1394
1137
# when a pack is being created by this object, the state of that pack.
1395
1138
self._new_pack = None
1396
1139
# aggregated revision index data
1397
flush = self._flush_new_pack
1398
self.revision_index = AggregateIndex(self.reload_pack_names, flush)
1399
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1400
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1401
self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1402
all_indices = [self.revision_index, self.inventory_index,
1403
self.text_index, self.signature_index]
1405
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1406
all_indices.append(self.chk_index)
1408
# used to determine if we're using a chk_index elsewhere.
1409
self.chk_index = None
1410
# Tell all the CombinedGraphIndex objects about each other, so they can
1411
# share hints about which pack names to search first.
1412
all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1413
for combined_idx in all_combined:
1414
combined_idx.set_sibling_indices(
1415
set(all_combined).difference([combined_idx]))
1417
self._resumed_packs = []
1420
return '%s(%r)' % (self.__class__.__name__, self.repo)
1140
self.revision_index = AggregateIndex()
1141
self.inventory_index = AggregateIndex()
1142
self.text_index = AggregateIndex()
1143
self.signature_index = AggregateIndex()
1422
1145
def add_pack_to_memory(self, pack):
1423
1146
"""Make a Pack object available to the repository to satisfy queries.
1425
1148
:param pack: A Pack object.
1427
if pack.name in self._packs_by_name:
1428
raise AssertionError(
1429
'pack %s already in _packs_by_name' % (pack.name,))
1150
assert pack.name not in self._packs_by_name
1430
1151
self.packs.append(pack)
1431
1152
self._packs_by_name[pack.name] = pack
1432
1153
self.revision_index.add_index(pack.revision_index, pack)
1433
1154
self.inventory_index.add_index(pack.inventory_index, pack)
1434
1155
self.text_index.add_index(pack.text_index, pack)
1435
1156
self.signature_index.add_index(pack.signature_index, pack)
1436
if self.chk_index is not None:
1437
self.chk_index.add_index(pack.chk_index, pack)
1158
def _add_text_to_weave(self, file_id, revision_id, new_lines, parents,
1159
nostore_sha, random_revid):
1160
file_id_index = GraphIndexPrefixAdapter(
1161
self.text_index.combined_index,
1163
add_nodes_callback=self.text_index.add_callback)
1164
self.repo._text_knit._index._graph_index = file_id_index
1165
self.repo._text_knit._index._add_callback = file_id_index.add_nodes
1166
return self.repo._text_knit.add_lines_with_ghosts(
1167
revision_id, parents, new_lines, nostore_sha=nostore_sha,
1168
random_id=random_revid, check_content=False)[0:2]
1439
1170
def all_packs(self):
1440
1171
"""Return a list of all the Pack objects this repository has.
1491
1218
# group their data with the relevant commit, and that may
1492
1219
# involve rewriting ancient history - which autopack tries to
1493
1220
# avoid. Alternatively we could not group the data but treat
1494
# each of these as having a single revision, and thus add
1221
# each of these as having a single revision, and thus add
1495
1222
# one revision for each to the total revision count, to get
1496
1223
# a matching distribution.
1498
1225
existing_packs.append((revision_count, pack))
1499
1226
pack_operations = self.plan_autopack_combinations(
1500
1227
existing_packs, pack_distribution)
1501
num_new_packs = len(pack_operations)
1502
num_old_packs = sum([len(po[1]) for po in pack_operations])
1503
num_revs_affected = sum([po[0] for po in pack_operations])
1504
mutter('Auto-packing repository %s, which has %d pack files, '
1505
'containing %d revisions. Packing %d files into %d affecting %d'
1506
' revisions', self, total_packs, total_revisions, num_old_packs,
1507
num_new_packs, num_revs_affected)
1508
result = self._execute_pack_operations(pack_operations,
1509
reload_func=self._restart_autopack)
1510
mutter('Auto-packing repository %s completed', self)
1228
self._execute_pack_operations(pack_operations)
1513
def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1231
def _execute_pack_operations(self, pack_operations, _packer_class=Packer):
1515
1232
"""Execute a series of pack operations.
1517
1234
:param pack_operations: A list of [revision_count, packs_to_combine].
1518
1235
:param _packer_class: The class of packer to use (default: Packer).
1519
:return: The new pack names.
1521
1238
for revision_count, packs in pack_operations:
1522
1239
# we may have no-ops from the setup logic
1523
1240
if len(packs) == 0:
1525
packer = _packer_class(self, packs, '.autopack',
1526
reload_func=reload_func)
1529
except errors.RetryWithNewPacks:
1530
# An exception is propagating out of this context, make sure
1531
# this packer has cleaned up. Packer() doesn't set its new_pack
1532
# state into the RepositoryPackCollection object, so we only
1533
# have access to it directly here.
1534
if packer.new_pack is not None:
1535
packer.new_pack.abort()
1242
_packer_class(self, packs, '.autopack').pack()
1537
1243
for pack in packs:
1538
1244
self._remove_pack_from_memory(pack)
1539
1245
# record the newly available packs and stop advertising the old
1541
to_be_obsoleted = []
1542
for _, packs in pack_operations:
1543
to_be_obsoleted.extend(packs)
1544
result = self._save_pack_names(clear_obsolete_packs=True,
1545
obsolete_packs=to_be_obsoleted)
1548
def _flush_new_pack(self):
1549
if self._new_pack is not None:
1550
self._new_pack.flush()
1247
self._save_pack_names(clear_obsolete_packs=True)
1248
# Move the old packs out of the way now they are no longer referenced.
1249
for revision_count, packs in pack_operations:
1250
self._obsolete_packs(packs)
1552
1252
def lock_names(self):
1553
1253
"""Acquire the mutex around the pack-names index.
1555
1255
This cannot be used in the middle of a read-only transaction on the
1558
1258
self.repo.control_files.lock_write()
1560
def _already_packed(self):
1561
"""Is the collection already packed?"""
1562
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1564
def pack(self, hint=None, clean_obsolete_packs=False):
1565
1261
"""Pack the pack collection totally."""
1566
1262
self.ensure_loaded()
1567
1263
total_packs = len(self._names)
1568
if self._already_packed():
1265
# This is arguably wrong because we might not be optimal, but for
1266
# now lets leave it in. (e.g. reconcile -> one pack. But not
1570
1269
total_revisions = self.revision_index.combined_index.key_count()
1571
1270
# XXX: the following may want to be a class, to pack with a given
1573
1272
mutter('Packing repository %s, which has %d pack files, '
1574
'containing %d revisions with hint %r.', self, total_packs,
1575
total_revisions, hint)
1273
'containing %d revisions into 1 packs.', self, total_packs,
1576
1275
# determine which packs need changing
1276
pack_distribution = [1]
1577
1277
pack_operations = [[0, []]]
1578
1278
for pack in self.all_packs():
1579
if hint is None or pack.name in hint:
1580
# Either no hint was provided (so we are packing everything),
1581
# or this pack was included in the hint.
1582
pack_operations[-1][0] += pack.get_revision_count()
1583
pack_operations[-1][1].append(pack)
1279
pack_operations[-1][0] += pack.get_revision_count()
1280
pack_operations[-1][1].append(pack)
1584
1281
self._execute_pack_operations(pack_operations, OptimisingPacker)
1586
if clean_obsolete_packs:
1587
self._clear_obsolete_packs()
1589
1283
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1590
1284
"""Plan a pack operation.
1874
1487
self._packs_by_name = {}
1875
1488
self._packs_at_load = None
1490
def _make_index_map(self, index_suffix):
1491
"""Return information on existing indices.
1493
:param suffix: Index suffix added to pack name.
1495
:returns: (pack_map, indices) where indices is a list of GraphIndex
1496
objects, and pack_map is a mapping from those objects to the
1497
pack tuple they describe.
1499
# TODO: stop using this; it creates new indices unnecessarily.
1500
self.ensure_loaded()
1501
suffix_map = {'.rix': 'revision_index',
1502
'.six': 'signature_index',
1503
'.iix': 'inventory_index',
1504
'.tix': 'text_index',
1506
return self._packs_list_to_pack_map_and_index_list(self.all_packs(),
1507
suffix_map[index_suffix])
1509
def _packs_list_to_pack_map_and_index_list(self, packs, index_attribute):
1510
"""Convert a list of packs to an index pack map and index list.
1512
:param packs: The packs list to process.
1513
:param index_attribute: The attribute that the desired index is found
1515
:return: A tuple (map, list) where map contains the dict from
1516
index:pack_tuple, and lsit contains the indices in the same order
1522
index = getattr(pack, index_attribute)
1523
indices.append(index)
1524
pack_map[index] = (pack.pack_transport, pack.file_name())
1525
return pack_map, indices
1527
def _index_contents(self, pack_map, key_filter=None):
1528
"""Get an iterable of the index contents from a pack_map.
1530
:param pack_map: A map from indices to pack details.
1531
:param key_filter: An optional filter to limit the
1534
indices = [index for index in pack_map.iterkeys()]
1535
all_index = CombinedGraphIndex(indices)
1536
if key_filter is None:
1537
return all_index.iter_all_entries()
1539
return all_index.iter_entries(key_filter)
1877
1541
def _unlock_names(self):
1878
1542
"""Release the mutex around the pack-names index."""
1879
1543
self.repo.control_files.unlock()
1881
def _diff_pack_names(self):
1882
"""Read the pack names from disk, and compare it to the one in memory.
1884
:return: (disk_nodes, deleted_nodes, new_nodes)
1885
disk_nodes The final set of nodes that should be referenced
1886
deleted_nodes Nodes which have been removed from when we started
1887
new_nodes Nodes that are newly introduced
1889
# load the disk nodes across
1891
for index, key, value in self._iter_disk_pack_index():
1892
disk_nodes.add((key, value))
1893
orig_disk_nodes = set(disk_nodes)
1895
# do a two-way diff against our original content
1896
current_nodes = set()
1897
for name, sizes in self._names.iteritems():
1899
((name, ), ' '.join(str(size) for size in sizes)))
1901
# Packs no longer present in the repository, which were present when we
1902
# locked the repository
1903
deleted_nodes = self._packs_at_load - current_nodes
1904
# Packs which this process is adding
1905
new_nodes = current_nodes - self._packs_at_load
1907
# Update the disk_nodes set to include the ones we are adding, and
1908
# remove the ones which were removed by someone else
1909
disk_nodes.difference_update(deleted_nodes)
1910
disk_nodes.update(new_nodes)
1912
return disk_nodes, deleted_nodes, new_nodes, orig_disk_nodes
1914
def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1915
"""Given the correct set of pack files, update our saved info.
1917
:return: (removed, added, modified)
1918
removed pack names removed from self._names
1919
added pack names added to self._names
1920
modified pack names that had changed value
1925
## self._packs_at_load = disk_nodes
1545
def _save_pack_names(self, clear_obsolete_packs=False):
1546
"""Save the list of packs.
1548
This will take out the mutex around the pack names list for the
1549
duration of the method call. If concurrent updates have been made, a
1550
three-way merge between the current list and the current in memory list
1553
:param clear_obsolete_packs: If True, clear out the contents of the
1554
obsolete_packs directory.
1558
builder = GraphIndexBuilder()
1559
# load the disk nodes across
1561
for index, key, value in self._iter_disk_pack_index():
1562
disk_nodes.add((key, value))
1563
# do a two-way diff against our original content
1564
current_nodes = set()
1565
for name, sizes in self._names.iteritems():
1567
((name, ), ' '.join(str(size) for size in sizes)))
1568
deleted_nodes = self._packs_at_load - current_nodes
1569
new_nodes = current_nodes - self._packs_at_load
1570
disk_nodes.difference_update(deleted_nodes)
1571
disk_nodes.update(new_nodes)
1572
# TODO: handle same-name, index-size-changes here -
1573
# e.g. use the value from disk, not ours, *unless* we're the one
1575
for key, value in disk_nodes:
1576
builder.add_node(key, value)
1577
self.transport.put_file('pack-names', builder.finish(),
1578
mode=self.repo.control_files._file_mode)
1579
# move the baseline forward
1580
self._packs_at_load = disk_nodes
1581
# now clear out the obsolete packs directory
1582
if clear_obsolete_packs:
1583
self.transport.clone('obsolete_packs').delete_multi(
1584
self.transport.list_dir('obsolete_packs'))
1586
self._unlock_names()
1587
# synchronise the memory packs list with what we just wrote:
1926
1588
new_names = dict(disk_nodes)
1927
1589
# drop no longer present nodes
1928
1590
for pack in self.all_packs():
1929
1591
if (pack.name,) not in new_names:
1930
removed.append(pack.name)
1931
1592
self._remove_pack_from_memory(pack)
1932
1593
# add new nodes/refresh existing ones
1933
1594
for key, value in disk_nodes:
1947
1608
self._remove_pack_from_memory(self.get_pack_by_name(name))
1948
1609
self._names[name] = sizes
1949
1610
self.get_pack_by_name(name)
1950
modified.append(name)
1953
1613
self._names[name] = sizes
1954
1614
self.get_pack_by_name(name)
1956
return removed, added, modified
1958
def _save_pack_names(self, clear_obsolete_packs=False, obsolete_packs=None):
1959
"""Save the list of packs.
1961
This will take out the mutex around the pack names list for the
1962
duration of the method call. If concurrent updates have been made, a
1963
three-way merge between the current list and the current in memory list
1966
:param clear_obsolete_packs: If True, clear out the contents of the
1967
obsolete_packs directory.
1968
:param obsolete_packs: Packs that are obsolete once the new pack-names
1969
file has been written.
1970
:return: A list of the names saved that were not previously on disk.
1972
already_obsolete = []
1975
builder = self._index_builder_class()
1976
(disk_nodes, deleted_nodes, new_nodes,
1977
orig_disk_nodes) = self._diff_pack_names()
1978
# TODO: handle same-name, index-size-changes here -
1979
# e.g. use the value from disk, not ours, *unless* we're the one
1981
for key, value in disk_nodes:
1982
builder.add_node(key, value)
1983
self.transport.put_file('pack-names', builder.finish(),
1984
mode=self.repo.bzrdir._get_file_mode())
1985
self._packs_at_load = disk_nodes
1986
if clear_obsolete_packs:
1989
to_preserve = set([o.name for o in obsolete_packs])
1990
already_obsolete = self._clear_obsolete_packs(to_preserve)
1992
self._unlock_names()
1993
# synchronise the memory packs list with what we just wrote:
1994
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1996
# TODO: We could add one more condition here. "if o.name not in
1997
# orig_disk_nodes and o != the new_pack we haven't written to
1998
# disk yet. However, the new pack object is not easily
1999
# accessible here (it would have to be passed through the
2000
# autopacking code, etc.)
2001
obsolete_packs = [o for o in obsolete_packs
2002
if o.name not in already_obsolete]
2003
self._obsolete_packs(obsolete_packs)
2004
return [new_node[0][0] for new_node in new_nodes]
2006
def reload_pack_names(self):
2007
"""Sync our pack listing with what is present in the repository.
2009
This should be called when we find out that something we thought was
2010
present is now missing. This happens when another process re-packs the
2013
:return: True if the in-memory list of packs has been altered at all.
2015
# The ensure_loaded call is to handle the case where the first call
2016
# made involving the collection was to reload_pack_names, where we
2017
# don't have a view of disk contents. Its a bit of a bandaid, and
2018
# causes two reads of pack-names, but its a rare corner case not struck
2019
# with regular push/pull etc.
2020
first_read = self.ensure_loaded()
2023
# out the new value.
2024
(disk_nodes, deleted_nodes, new_nodes,
2025
orig_disk_nodes) = self._diff_pack_names()
2026
# _packs_at_load is meant to be the explicit list of names in
2027
# 'pack-names' at then start. As such, it should not contain any
2028
# pending names that haven't been written out yet.
2029
self._packs_at_load = orig_disk_nodes
2031
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2032
if removed or added or modified:
2036
def _restart_autopack(self):
2037
"""Reload the pack names list, and restart the autopack code."""
2038
if not self.reload_pack_names():
2039
# Re-raise the original exception, because something went missing
2040
# and a restart didn't find it
2042
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2044
def _clear_obsolete_packs(self, preserve=None):
2045
"""Delete everything from the obsolete-packs directory.
2047
:return: A list of pack identifiers (the filename without '.pack') that
2048
were found in obsolete_packs.
2051
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2052
if preserve is None:
2054
for filename in obsolete_pack_transport.list_dir('.'):
2055
name, ext = osutils.splitext(filename)
2058
if name in preserve:
2061
obsolete_pack_transport.delete(filename)
2062
except (errors.PathError, errors.TransportError), e:
2063
warning("couldn't delete obsolete pack, skipping it:\n%s"
2067
1616
def _start_write_group(self):
2068
1617
# Do not permit preparation for writing if we're not in a 'write lock'.
2069
1618
if not self.repo.is_write_locked():
2070
1619
raise errors.NotWriteLocked(self)
2071
self._new_pack = self.pack_factory(self, upload_suffix='.pack',
2072
file_mode=self.repo.bzrdir._get_file_mode())
1620
self._new_pack = NewPack(self._upload_transport, self._index_transport,
1621
self._pack_transport, upload_suffix='.pack',
1622
file_mode=self.repo.control_files._file_mode)
2073
1623
# allow writing: queue writes to a new index
2074
1624
self.revision_index.add_writable_index(self._new_pack.revision_index,
2075
1625
self._new_pack)
2077
1627
self._new_pack)
2078
1628
self.text_index.add_writable_index(self._new_pack.text_index,
2079
1629
self._new_pack)
2080
self._new_pack.text_index.set_optimize(combine_backing_indices=False)
2081
1630
self.signature_index.add_writable_index(self._new_pack.signature_index,
2082
1631
self._new_pack)
2083
if self.chk_index is not None:
2084
self.chk_index.add_writable_index(self._new_pack.chk_index,
2086
self.repo.chk_bytes._index._add_callback = self.chk_index.add_callback
2087
self._new_pack.chk_index.set_optimize(combine_backing_indices=False)
2089
self.repo.inventories._index._add_callback = self.inventory_index.add_callback
2090
self.repo.revisions._index._add_callback = self.revision_index.add_callback
2091
self.repo.signatures._index._add_callback = self.signature_index.add_callback
2092
self.repo.texts._index._add_callback = self.text_index.add_callback
1633
# reused revision and signature knits may need updating
1635
# "Hysterical raisins. client code in bzrlib grabs those knits outside
1636
# of write groups and then mutates it inside the write group."
1637
if self.repo._revision_knit is not None:
1638
self.repo._revision_knit._index._add_callback = \
1639
self.revision_index.add_callback
1640
if self.repo._signature_knit is not None:
1641
self.repo._signature_knit._index._add_callback = \
1642
self.signature_index.add_callback
1643
# create a reused knit object for text addition in commit.
1644
self.repo._text_knit = self.repo.weave_store.get_weave_or_empty(
2094
1647
def _abort_write_group(self):
2095
1648
# FIXME: just drop the transient index.
2096
1649
# forget what names there are
2097
if self._new_pack is not None:
2098
operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2099
operation.add_cleanup(setattr, self, '_new_pack', None)
2100
# If we aborted while in the middle of finishing the write
2101
# group, _remove_pack_indices could fail because the indexes are
2102
# already gone. But they're not there we shouldn't fail in this
2103
# case, so we pass ignore_missing=True.
2104
operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2105
ignore_missing=True)
2106
operation.run_simple()
2107
for resumed_pack in self._resumed_packs:
2108
operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2109
# See comment in previous finally block.
2110
operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2111
ignore_missing=True)
2112
operation.run_simple()
2113
del self._resumed_packs[:]
2115
def _remove_resumed_pack_indices(self):
2116
for resumed_pack in self._resumed_packs:
2117
self._remove_pack_indices(resumed_pack)
2118
del self._resumed_packs[:]
2120
def _check_new_inventories(self):
2121
"""Detect missing inventories in this write group.
2123
:returns: list of strs, summarising any problems found. If the list is
2124
empty no problems were found.
2126
# The base implementation does no checks. GCRepositoryPackCollection
1650
self._new_pack.abort()
1651
self._remove_pack_indices(self._new_pack)
1652
self._new_pack = None
1653
self.repo._text_knit = None
2130
1655
def _commit_write_group(self):
2132
for prefix, versioned_file in (
2133
('revisions', self.repo.revisions),
2134
('inventories', self.repo.inventories),
2135
('texts', self.repo.texts),
2136
('signatures', self.repo.signatures),
2138
missing = versioned_file.get_missing_compression_parent_keys()
2139
all_missing.update([(prefix,) + key for key in missing])
2141
raise errors.BzrCheckError(
2142
"Repository %s has missing compression parent(s) %r "
2143
% (self.repo, sorted(all_missing)))
2144
problems = self._check_new_inventories()
2146
problems_summary = '\n'.join(problems)
2147
raise errors.BzrCheckError(
2148
"Cannot add revision(s) to repository: " + problems_summary)
2149
1656
self._remove_pack_indices(self._new_pack)
2150
any_new_content = False
2151
1657
if self._new_pack.data_inserted():
2152
1658
# get all the data to disk and read to use
2153
1659
self._new_pack.finish()
2154
1660
self.allocate(self._new_pack)
2155
1661
self._new_pack = None
2156
any_new_content = True
2158
self._new_pack.abort()
2159
self._new_pack = None
2160
for resumed_pack in self._resumed_packs:
2161
# XXX: this is a pretty ugly way to turn the resumed pack into a
2162
# properly committed pack.
2163
self._names[resumed_pack.name] = None
2164
self._remove_pack_from_memory(resumed_pack)
2165
resumed_pack.finish()
2166
self.allocate(resumed_pack)
2167
any_new_content = True
2168
del self._resumed_packs[:]
2170
result = self.autopack()
1662
if not self.autopack():
2172
1663
# when autopack takes no steps, the names list is still
2174
return self._save_pack_names()
2178
def _suspend_write_group(self):
2179
tokens = [pack.name for pack in self._resumed_packs]
2180
self._remove_pack_indices(self._new_pack)
2181
if self._new_pack.data_inserted():
2182
# get all the data to disk and read to use
2183
self._new_pack.finish(suspend=True)
2184
tokens.append(self._new_pack.name)
2185
self._new_pack = None
1665
self._save_pack_names()
2187
1667
self._new_pack.abort()
2188
1668
self._new_pack = None
2189
self._remove_resumed_pack_indices()
2192
def _resume_write_group(self, tokens):
2193
for token in tokens:
2194
self._resume_pack(token)
1669
self.repo._text_knit = None
1672
class KnitPackRevisionStore(KnitRevisionStore):
1673
"""An object to adapt access from RevisionStore's to use KnitPacks.
1675
This class works by replacing the original RevisionStore.
1676
We need to do this because the KnitPackRevisionStore is less
1677
isolated in its layering - it uses services from the repo.
1680
def __init__(self, repo, transport, revisionstore):
1681
"""Create a KnitPackRevisionStore on repo with revisionstore.
1683
This will store its state in the Repository, use the
1684
indices to provide a KnitGraphIndex,
1685
and at the end of transactions write new indices.
1687
KnitRevisionStore.__init__(self, revisionstore.versioned_file_store)
1689
self._serializer = revisionstore._serializer
1690
self.transport = transport
1692
def get_revision_file(self, transaction):
1693
"""Get the revision versioned file object."""
1694
if getattr(self.repo, '_revision_knit', None) is not None:
1695
return self.repo._revision_knit
1696
self.repo._pack_collection.ensure_loaded()
1697
add_callback = self.repo._pack_collection.revision_index.add_callback
1698
# setup knit specific objects
1699
knit_index = KnitGraphIndex(
1700
self.repo._pack_collection.revision_index.combined_index,
1701
add_callback=add_callback)
1702
self.repo._revision_knit = knit.KnitVersionedFile(
1703
'revisions', self.transport.clone('..'),
1704
self.repo.control_files._file_mode,
1705
create=False, access_mode=self.repo._access_mode(),
1706
index=knit_index, delta=False, factory=knit.KnitPlainFactory(),
1707
access_method=self.repo._pack_collection.revision_index.knit_access)
1708
return self.repo._revision_knit
1710
def get_signature_file(self, transaction):
1711
"""Get the signature versioned file object."""
1712
if getattr(self.repo, '_signature_knit', None) is not None:
1713
return self.repo._signature_knit
1714
self.repo._pack_collection.ensure_loaded()
1715
add_callback = self.repo._pack_collection.signature_index.add_callback
1716
# setup knit specific objects
1717
knit_index = KnitGraphIndex(
1718
self.repo._pack_collection.signature_index.combined_index,
1719
add_callback=add_callback, parents=False)
1720
self.repo._signature_knit = knit.KnitVersionedFile(
1721
'signatures', self.transport.clone('..'),
1722
self.repo.control_files._file_mode,
1723
create=False, access_mode=self.repo._access_mode(),
1724
index=knit_index, delta=False, factory=knit.KnitPlainFactory(),
1725
access_method=self.repo._pack_collection.signature_index.knit_access)
1726
return self.repo._signature_knit
1729
class KnitPackTextStore(VersionedFileStore):
1730
"""Presents a TextStore abstraction on top of packs.
1732
This class works by replacing the original VersionedFileStore.
1733
We need to do this because the KnitPackRevisionStore is less
1734
isolated in its layering - it uses services from the repo and shares them
1735
with all the data written in a single write group.
1738
def __init__(self, repo, transport, weavestore):
1739
"""Create a KnitPackTextStore on repo with weavestore.
1741
This will store its state in the Repository, use the
1742
indices FileNames to provide a KnitGraphIndex,
1743
and at the end of transactions write new indices.
1745
# don't call base class constructor - it's not suitable.
1746
# no transient data stored in the transaction
1748
self._precious = False
1750
self.transport = transport
1751
self.weavestore = weavestore
1752
# XXX for check() which isn't updated yet
1753
self._transport = weavestore._transport
1755
def get_weave_or_empty(self, file_id, transaction):
1756
"""Get a 'Knit' backed by the .tix indices.
1758
The transaction parameter is ignored.
1760
self.repo._pack_collection.ensure_loaded()
1761
add_callback = self.repo._pack_collection.text_index.add_callback
1762
# setup knit specific objects
1763
file_id_index = GraphIndexPrefixAdapter(
1764
self.repo._pack_collection.text_index.combined_index,
1765
(file_id, ), 1, add_nodes_callback=add_callback)
1766
knit_index = KnitGraphIndex(file_id_index,
1767
add_callback=file_id_index.add_nodes,
1768
deltas=True, parents=True)
1769
return knit.KnitVersionedFile('text:' + file_id,
1770
self.transport.clone('..'),
1773
access_method=self.repo._pack_collection.text_index.knit_access,
1774
factory=knit.KnitPlainFactory())
1776
get_weave = get_weave_or_empty
1779
"""Generate a list of the fileids inserted, for use by check."""
1780
self.repo._pack_collection.ensure_loaded()
1782
for index, key, value, refs in \
1783
self.repo._pack_collection.text_index.combined_index.iter_all_entries():
1788
class InventoryKnitThunk(object):
1789
"""An object to manage thunking get_inventory_weave to pack based knits."""
1791
def __init__(self, repo, transport):
1792
"""Create an InventoryKnitThunk for repo at transport.
1794
This will store its state in the Repository, use the
1795
indices FileNames to provide a KnitGraphIndex,
1796
and at the end of transactions write a new index..
1799
self.transport = transport
1801
def get_weave(self):
1802
"""Get a 'Knit' that contains inventory data."""
1803
self.repo._pack_collection.ensure_loaded()
1804
add_callback = self.repo._pack_collection.inventory_index.add_callback
1805
# setup knit specific objects
1806
knit_index = KnitGraphIndex(
1807
self.repo._pack_collection.inventory_index.combined_index,
1808
add_callback=add_callback, deltas=True, parents=True)
1809
return knit.KnitVersionedFile(
1810
'inventory', self.transport.clone('..'),
1811
self.repo.control_files._file_mode,
1812
create=False, access_mode=self.repo._access_mode(),
1813
index=knit_index, delta=True, factory=knit.KnitPlainFactory(),
1814
access_method=self.repo._pack_collection.inventory_index.knit_access)
2197
1817
class KnitPackRepository(KnitRepository):
2198
"""Repository with knit objects stored inside pack containers.
2200
The layering for a KnitPackRepository is:
2202
Graph | HPSS | Repository public layer |
2203
===================================================
2204
Tuple based apis below, string based, and key based apis above
2205
---------------------------------------------------
2207
Provides .texts, .revisions etc
2208
This adapts the N-tuple keys to physical knit records which only have a
2209
single string identifier (for historical reasons), which in older formats
2210
was always the revision_id, and in the mapped code for packs is always
2211
the last element of key tuples.
2212
---------------------------------------------------
2214
A separate GraphIndex is used for each of the
2215
texts/inventories/revisions/signatures contained within each individual
2216
pack file. The GraphIndex layer works in N-tuples and is unaware of any
2218
===================================================
2222
def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
1818
"""Experimental graph-knit using repository."""
1820
def __init__(self, _format, a_bzrdir, control_files, _revision_store,
1821
control_store, text_store, _commit_builder_class, _serializer):
2224
1822
KnitRepository.__init__(self, _format, a_bzrdir, control_files,
2225
_commit_builder_class, _serializer)
2226
index_transport = self._transport.clone('indices')
2227
self._pack_collection = RepositoryPackCollection(self, self._transport,
1823
_revision_store, control_store, text_store, _commit_builder_class,
1825
index_transport = control_files._transport.clone('indices')
1826
self._pack_collection = RepositoryPackCollection(self, control_files._transport,
2228
1827
index_transport,
2229
self._transport.clone('upload'),
2230
self._transport.clone('packs'),
2231
_format.index_builder_class,
2232
_format.index_class,
2233
use_chk_index=self._format.supports_chks,
2235
self.inventories = KnitVersionedFiles(
2236
_KnitGraphIndex(self._pack_collection.inventory_index.combined_index,
2237
add_callback=self._pack_collection.inventory_index.add_callback,
2238
deltas=True, parents=True, is_locked=self.is_locked),
2239
data_access=self._pack_collection.inventory_index.data_access,
2240
max_delta_chain=200)
2241
self.revisions = KnitVersionedFiles(
2242
_KnitGraphIndex(self._pack_collection.revision_index.combined_index,
2243
add_callback=self._pack_collection.revision_index.add_callback,
2244
deltas=False, parents=True, is_locked=self.is_locked,
2245
track_external_parent_refs=True),
2246
data_access=self._pack_collection.revision_index.data_access,
2248
self.signatures = KnitVersionedFiles(
2249
_KnitGraphIndex(self._pack_collection.signature_index.combined_index,
2250
add_callback=self._pack_collection.signature_index.add_callback,
2251
deltas=False, parents=False, is_locked=self.is_locked),
2252
data_access=self._pack_collection.signature_index.data_access,
2254
self.texts = KnitVersionedFiles(
2255
_KnitGraphIndex(self._pack_collection.text_index.combined_index,
2256
add_callback=self._pack_collection.text_index.add_callback,
2257
deltas=True, parents=True, is_locked=self.is_locked),
2258
data_access=self._pack_collection.text_index.data_access,
2259
max_delta_chain=200)
2260
if _format.supports_chks:
2261
# No graph, no compression:- references from chks are between
2262
# different objects not temporal versions of the same; and without
2263
# some sort of temporal structure knit compression will just fail.
2264
self.chk_bytes = KnitVersionedFiles(
2265
_KnitGraphIndex(self._pack_collection.chk_index.combined_index,
2266
add_callback=self._pack_collection.chk_index.add_callback,
2267
deltas=False, parents=False, is_locked=self.is_locked),
2268
data_access=self._pack_collection.chk_index.data_access,
2271
self.chk_bytes = None
1828
control_files._transport.clone('upload'),
1829
control_files._transport.clone('packs'))
1830
self._revision_store = KnitPackRevisionStore(self, index_transport, self._revision_store)
1831
self.weave_store = KnitPackTextStore(self, index_transport, self.weave_store)
1832
self._inv_thunk = InventoryKnitThunk(self, index_transport)
2272
1833
# True when the repository object is 'write locked' (as opposed to the
2273
# physical lock only taken out around changes to the pack-names list.)
1834
# physical lock only taken out around changes to the pack-names list.)
2274
1835
# Another way to represent this would be a decorator around the control
2275
1836
# files object that presents logical locks as physical ones - if this
2276
1837
# gets ugly consider that alternative design. RBC 20071011
2281
1842
self._reconcile_fixes_text_parents = True
2282
1843
self._reconcile_backsup_inventory = False
2284
def _warn_if_deprecated(self, branch=None):
2285
# This class isn't deprecated, but one sub-format is
2286
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2287
super(KnitPackRepository, self)._warn_if_deprecated(branch)
2289
1845
def _abort_write_group(self):
2290
self.revisions._index._key_dependencies.clear()
2291
1846
self._pack_collection._abort_write_group()
2293
def _get_source(self, to_format):
2294
if to_format.network_name() == self._format.network_name():
2295
return KnitPackStreamSource(self, to_format)
2296
return super(KnitPackRepository, self)._get_source(to_format)
1848
def _access_mode(self):
1849
"""Return 'w' or 'r' for depending on whether a write lock is active.
1851
This method is a helper for the Knit-thunking support objects.
1853
if self.is_write_locked():
1857
def _find_inconsistent_revision_parents(self):
1858
"""Find revisions with incorrectly cached parents.
1860
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1861
parents-in-revision).
1863
if not self.is_locked():
1864
raise errors.ObjectNotLocked(self)
1865
pb = ui.ui_factory.nested_progress_bar()
1868
revision_nodes = self._pack_collection.revision_index \
1869
.combined_index.iter_all_entries()
1870
index_positions = []
1871
# Get the cached index values for all revisions, and also the location
1872
# in each index of the revision text so we can perform linear IO.
1873
for index, key, value, refs in revision_nodes:
1874
pos, length = value[1:].split(' ')
1875
index_positions.append((index, int(pos), key[0],
1876
tuple(parent[0] for parent in refs[0])))
1877
pb.update("Reading revision index.", 0, 0)
1878
index_positions.sort()
1879
batch_count = len(index_positions) / 1000 + 1
1880
pb.update("Checking cached revision graph.", 0, batch_count)
1881
for offset in xrange(batch_count):
1882
pb.update("Checking cached revision graph.", offset)
1883
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
1886
rev_ids = [item[2] for item in to_query]
1887
revs = self.get_revisions(rev_ids)
1888
for revision, item in zip(revs, to_query):
1889
index_parents = item[3]
1890
rev_parents = tuple(revision.parent_ids)
1891
if index_parents != rev_parents:
1892
result.append((revision.revision_id, index_parents, rev_parents))
1897
def get_parents(self, revision_ids):
1898
"""See StackedParentsProvider.get_parents.
1900
This implementation accesses the combined revision index to provide
1903
self._pack_collection.ensure_loaded()
1904
index = self._pack_collection.revision_index.combined_index
1906
for revision_id in revision_ids:
1907
if revision_id != _mod_revision.NULL_REVISION:
1908
search_keys.add((revision_id,))
1909
found_parents = {_mod_revision.NULL_REVISION:[]}
1910
for index, key, value, refs in index.iter_entries(search_keys):
1913
parents = (_mod_revision.NULL_REVISION,)
1915
parents = tuple(parent[0] for parent in parents)
1916
found_parents[key[0]] = parents
1918
for revision_id in revision_ids:
1920
result.append(found_parents[revision_id])
2298
1925
def _make_parents_provider(self):
2299
return graph.CachingParentsProvider(self)
2301
1928
def _refresh_data(self):
2302
if not self.is_locked():
2304
self._pack_collection.reload_pack_names()
1929
if self._write_lock_count == 1 or (
1930
self.control_files._lock_count == 1 and
1931
self.control_files._lock_mode == 'r'):
1932
# forget what names there are
1933
self._pack_collection.reset()
1934
# XXX: Better to do an in-memory merge when acquiring a new lock -
1935
# factor out code from _save_pack_names.
1936
self._pack_collection.ensure_loaded()
2306
1938
def _start_write_group(self):
2307
1939
self._pack_collection._start_write_group()
2309
1941
def _commit_write_group(self):
2310
hint = self._pack_collection._commit_write_group()
2311
self.revisions._index._key_dependencies.clear()
2314
def suspend_write_group(self):
2315
# XXX check self._write_group is self.get_transaction()?
2316
tokens = self._pack_collection._suspend_write_group()
2317
self.revisions._index._key_dependencies.clear()
2318
self._write_group = None
2321
def _resume_write_group(self, tokens):
2322
self._start_write_group()
2324
self._pack_collection._resume_write_group(tokens)
2325
except errors.UnresumableWriteGroup:
2326
self._abort_write_group()
2328
for pack in self._pack_collection._resumed_packs:
2329
self.revisions._index.scan_unvalidated_index(pack.revision_index)
1942
return self._pack_collection._commit_write_group()
1944
def get_inventory_weave(self):
1945
return self._inv_thunk.get_weave()
2331
1947
def get_transaction(self):
2332
1948
if self._write_lock_count:
2420
2015
self.control_files.unlock()
2422
if not self.is_locked():
2423
for repo in self._fallback_repositories:
2427
class KnitPackStreamSource(StreamSource):
2428
"""A StreamSource used to transfer data between same-format KnitPack repos.
2430
This source assumes:
2431
1) Same serialization format for all objects
2432
2) Same root information
2433
3) XML format inventories
2434
4) Atomic inserts (so we can stream inventory texts before text
2439
def __init__(self, from_repository, to_format):
2440
super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2441
self._text_keys = None
2442
self._text_fetch_order = 'unordered'
2444
def _get_filtered_inv_stream(self, revision_ids):
2445
from_repo = self.from_repository
2446
parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2447
parent_keys = [(p,) for p in parent_ids]
2448
find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2449
parent_text_keys = set(find_text_keys(
2450
from_repo._inventory_xml_lines_for_keys(parent_keys)))
2451
content_text_keys = set()
2452
knit = KnitVersionedFiles(None, None)
2453
factory = KnitPlainFactory()
2454
def find_text_keys_from_content(record):
2455
if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2456
raise ValueError("Unknown content storage kind for"
2457
" inventory text: %s" % (record.storage_kind,))
2458
# It's a knit record, it has a _raw_record field (even if it was
2459
# reconstituted from a network stream).
2460
raw_data = record._raw_record
2461
# read the entire thing
2462
revision_id = record.key[-1]
2463
content, _ = knit._parse_record(revision_id, raw_data)
2464
if record.storage_kind == 'knit-delta-gz':
2465
line_iterator = factory.get_linedelta_content(content)
2466
elif record.storage_kind == 'knit-ft-gz':
2467
line_iterator = factory.get_fulltext_content(content)
2468
content_text_keys.update(find_text_keys(
2469
[(line, revision_id) for line in line_iterator]))
2470
revision_keys = [(r,) for r in revision_ids]
2471
def _filtered_inv_stream():
2472
source_vf = from_repo.inventories
2473
stream = source_vf.get_record_stream(revision_keys,
2475
for record in stream:
2476
if record.storage_kind == 'absent':
2477
raise errors.NoSuchRevision(from_repo, record.key)
2478
find_text_keys_from_content(record)
2480
self._text_keys = content_text_keys - parent_text_keys
2481
return ('inventories', _filtered_inv_stream())
2483
def _get_text_stream(self):
2484
# Note: We know we don't have to handle adding root keys, because both
2485
# the source and target are the identical network name.
2486
text_stream = self.from_repository.texts.get_record_stream(
2487
self._text_keys, self._text_fetch_order, False)
2488
return ('texts', text_stream)
2490
def get_stream(self, search):
2491
revision_ids = search.get_keys()
2492
for stream_info in self._fetch_revision_texts(revision_ids):
2494
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2495
yield self._get_filtered_inv_stream(revision_ids)
2496
yield self._get_text_stream()
2500
2018
class RepositoryFormatPack(MetaDirRepositoryFormat):
2501
2019
"""Format logic for pack structured repositories.
2684
2234
def get_format_description(self):
2685
2235
"""See RepositoryFormat.get_format_description()."""
2686
2236
return "Packs containing knits with rich root support\n"
2689
class RepositoryFormatKnitPack5(RepositoryFormatPack):
2690
"""Repository that supports external references to allow stacking.
2694
Supports external lookups, which results in non-truncated ghosts after
2695
reconcile compared to pack-0.92 formats.
2698
repository_class = KnitPackRepository
2699
_commit_builder_class = PackCommitBuilder
2700
supports_external_lookups = True
2701
# What index classes to use
2702
index_builder_class = InMemoryGraphIndex
2703
index_class = GraphIndex
2706
def _serializer(self):
2707
return xml5.serializer_v5
2709
def _get_matching_bzrdir(self):
2710
return bzrdir.format_registry.make_bzrdir('1.6')
2712
def _ignore_setting_bzrdir(self, format):
2715
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2717
def get_format_string(self):
2718
"""See RepositoryFormat.get_format_string()."""
2719
return "Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n"
2721
def get_format_description(self):
2722
"""See RepositoryFormat.get_format_description()."""
2723
return "Packs 5 (adds stacking support, requires bzr 1.6)"
2726
class RepositoryFormatKnitPack5RichRoot(RepositoryFormatPack):
2727
"""A repository with rich roots and stacking.
2729
New in release 1.6.1.
2731
Supports stacking on other repositories, allowing data to be accessed
2732
without being stored locally.
2735
repository_class = KnitPackRepository
2736
_commit_builder_class = PackRootCommitBuilder
2737
rich_root_data = True
2738
supports_tree_reference = False # no subtrees
2739
supports_external_lookups = True
2740
# What index classes to use
2741
index_builder_class = InMemoryGraphIndex
2742
index_class = GraphIndex
2745
def _serializer(self):
2746
return xml6.serializer_v6
2748
def _get_matching_bzrdir(self):
2749
return bzrdir.format_registry.make_bzrdir(
2752
def _ignore_setting_bzrdir(self, format):
2755
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2757
def get_format_string(self):
2758
"""See RepositoryFormat.get_format_string()."""
2759
return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n"
2761
def get_format_description(self):
2762
return "Packs 5 rich-root (adds stacking support, requires bzr 1.6.1)"
2765
class RepositoryFormatKnitPack5RichRootBroken(RepositoryFormatPack):
2766
"""A repository with rich roots and external references.
2770
Supports external lookups, which results in non-truncated ghosts after
2771
reconcile compared to pack-0.92 formats.
2773
This format was deprecated because the serializer it uses accidentally
2774
supported subtrees, when the format was not intended to. This meant that
2775
someone could accidentally fetch from an incorrect repository.
2778
repository_class = KnitPackRepository
2779
_commit_builder_class = PackRootCommitBuilder
2780
rich_root_data = True
2781
supports_tree_reference = False # no subtrees
2783
supports_external_lookups = True
2784
# What index classes to use
2785
index_builder_class = InMemoryGraphIndex
2786
index_class = GraphIndex
2789
def _serializer(self):
2790
return xml7.serializer_v7
2792
def _get_matching_bzrdir(self):
2793
matching = bzrdir.format_registry.make_bzrdir(
2795
matching.repository_format = self
2798
def _ignore_setting_bzrdir(self, format):
2801
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2803
def get_format_string(self):
2804
"""See RepositoryFormat.get_format_string()."""
2805
return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n"
2807
def get_format_description(self):
2808
return ("Packs 5 rich-root (adds stacking support, requires bzr 1.6)"
2812
class RepositoryFormatKnitPack6(RepositoryFormatPack):
2813
"""A repository with stacking and btree indexes,
2814
without rich roots or subtrees.
2816
This is equivalent to pack-1.6 with B+Tree indices.
2819
repository_class = KnitPackRepository
2820
_commit_builder_class = PackCommitBuilder
2821
supports_external_lookups = True
2822
# What index classes to use
2823
index_builder_class = BTreeBuilder
2824
index_class = BTreeGraphIndex
2827
def _serializer(self):
2828
return xml5.serializer_v5
2830
def _get_matching_bzrdir(self):
2831
return bzrdir.format_registry.make_bzrdir('1.9')
2833
def _ignore_setting_bzrdir(self, format):
2836
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2838
def get_format_string(self):
2839
"""See RepositoryFormat.get_format_string()."""
2840
return "Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n"
2842
def get_format_description(self):
2843
"""See RepositoryFormat.get_format_description()."""
2844
return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2847
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2848
"""A repository with rich roots, no subtrees, stacking and btree indexes.
2850
1.6-rich-root with B+Tree indices.
2853
repository_class = KnitPackRepository
2854
_commit_builder_class = PackRootCommitBuilder
2855
rich_root_data = True
2856
supports_tree_reference = False # no subtrees
2857
supports_external_lookups = True
2858
# What index classes to use
2859
index_builder_class = BTreeBuilder
2860
index_class = BTreeGraphIndex
2863
def _serializer(self):
2864
return xml6.serializer_v6
2866
def _get_matching_bzrdir(self):
2867
return bzrdir.format_registry.make_bzrdir(
2870
def _ignore_setting_bzrdir(self, format):
2873
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2875
def get_format_string(self):
2876
"""See RepositoryFormat.get_format_string()."""
2877
return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2879
def get_format_description(self):
2880
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2883
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2884
"""A subtrees development repository.
2886
This format should be retained until the second release after bzr 1.7.
2888
1.6.1-subtree[as it might have been] with B+Tree indices.
2890
This is [now] retained until we have a CHK based subtree format in
2894
repository_class = KnitPackRepository
2895
_commit_builder_class = PackRootCommitBuilder
2896
rich_root_data = True
2898
supports_tree_reference = True
2899
supports_external_lookups = True
2900
# What index classes to use
2901
index_builder_class = BTreeBuilder
2902
index_class = BTreeGraphIndex
2905
def _serializer(self):
2906
return xml7.serializer_v7
2908
def _get_matching_bzrdir(self):
2909
return bzrdir.format_registry.make_bzrdir(
2910
'development-subtree')
2912
def _ignore_setting_bzrdir(self, format):
2915
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2917
def get_format_string(self):
2918
"""See RepositoryFormat.get_format_string()."""
2919
return ("Bazaar development format 2 with subtree support "
2920
"(needs bzr.dev from before 1.8)\n")
2922
def get_format_description(self):
2923
"""See RepositoryFormat.get_format_description()."""
2924
return ("Development repository format, currently the same as "
2925
"1.6.1-subtree with B+Tree indices.\n")
added
removed
bzrlib/repofmt/pack_repo.py
