227
222
return self.index_name('text', name)
229
224
def _replace_index_with_readonly(self, index_type):
230
unlimited_cache = False
231
if index_type == 'chk':
232
unlimited_cache = True
233
225
setattr(self, index_type + '_index',
234
226
self.index_class(self.index_transport,
235
227
self.index_name(index_type, self.name),
236
self.index_sizes[self.index_offset(index_type)],
237
unlimited_cache=unlimited_cache))
228
self.index_sizes[self.index_offset(index_type)]))
240
231
class ExistingPack(Pack):
241
232
"""An in memory proxy for an existing .pack and its disk indices."""
243
234
def __init__(self, pack_transport, name, revision_index, inventory_index,
244
text_index, signature_index, chk_index=None):
235
text_index, signature_index):
245
236
"""Create an ExistingPack object.
247
238
:param pack_transport: The transport where the pack file resides.
248
239
:param name: The name of the pack on disk in the pack_transport.
250
241
Pack.__init__(self, revision_index, inventory_index, text_index,
251
signature_index, chk_index)
253
244
self.pack_transport = pack_transport
254
245
if None in (revision_index, inventory_index, text_index,
272
263
def __init__(self, name, revision_index, inventory_index, text_index,
273
264
signature_index, upload_transport, pack_transport, index_transport,
274
pack_collection, chk_index=None):
275
266
"""Create a ResumedPack object."""
276
267
ExistingPack.__init__(self, pack_transport, name, revision_index,
277
inventory_index, text_index, signature_index,
268
inventory_index, text_index, signature_index)
279
269
self.upload_transport = upload_transport
280
270
self.index_transport = index_transport
281
271
self.index_sizes = [None, None, None, None]
308
295
self.upload_transport.delete(self.file_name())
309
296
indices = [self.revision_index, self.inventory_index, self.text_index,
310
297
self.signature_index]
311
if self.chk_index is not None:
312
indices.append(self.chk_index)
313
298
for index in indices:
314
299
index._transport.delete(index._name)
316
301
def finish(self):
317
302
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:
303
new_name = '../packs/' + self.file_name()
304
self.upload_transport.rename(self.file_name(), new_name)
305
for index_type in ['revision', 'inventory', 'text', 'signature']:
322
306
old_name = self.index_name(index_type, self.name)
323
307
new_name = '../indices/' + old_name
324
308
self.upload_transport.rename(old_name, new_name)
325
309
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
310
self._state = 'finished'
330
312
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
313
return index.external_references(1)
584
543
self.index_to_pack = {}
585
544
self.combined_index = CombinedGraphIndex([], reload_func=reload_func)
586
545
self.data_access = _DirectPackAccess(self.index_to_pack,
587
reload_func=reload_func,
588
flush_func=flush_func)
546
reload_func=reload_func)
547
self.add_callback = None
549
def replace_indices(self, index_to_pack, indices):
550
"""Replace the current mappings with fresh ones.
552
This should probably not be used eventually, rather incremental add and
553
removal of indices. It has been added during refactoring of existing
556
:param index_to_pack: A mapping from index objects to
557
(transport, name) tuples for the pack file data.
558
:param indices: A list of indices.
560
# refresh the revision pack map dict without replacing the instance.
561
self.index_to_pack.clear()
562
self.index_to_pack.update(index_to_pack)
563
# XXX: API break - clearly a 'replace' method would be good?
564
self.combined_index._indices[:] = indices
565
# the current add nodes callback for the current writable index if
589
567
self.add_callback = None
591
569
def add_index(self, index, pack):
626
604
self.data_access.set_writer(None, None, (None, None))
627
605
self.index_to_pack.clear()
628
606
del self.combined_index._indices[:]
629
del self.combined_index._index_names[:]
630
607
self.add_callback = None
632
def remove_index(self, index):
609
def remove_index(self, index, pack):
633
610
"""Remove index from the indices used to answer queries.
635
612
:param index: An index from the pack parameter.
613
:param pack: A Pack instance.
637
615
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]
616
self.combined_index._indices.remove(index)
641
617
if (self.add_callback is not None and
642
618
getattr(index, 'add_nodes', None) == self.add_callback):
643
619
self.add_callback = None
750
726
def open_pack(self):
751
727
"""Open a pack for the pack we are creating."""
752
new_pack = self._pack_collection.pack_factory(self._pack_collection,
753
upload_suffix=self.suffix,
728
return NewPack(self._pack_collection, upload_suffix=self.suffix,
754
729
file_mode=self._pack_collection.repo.bzrdir._get_file_mode())
755
# We know that we will process all nodes in order, and don't need to
756
# query, so don't combine any indices spilled to disk until we are done
757
new_pack.revision_index.set_optimize(combine_backing_indices=False)
758
new_pack.inventory_index.set_optimize(combine_backing_indices=False)
759
new_pack.text_index.set_optimize(combine_backing_indices=False)
760
new_pack.signature_index.set_optimize(combine_backing_indices=False)
763
731
def _update_pack_order(self, entries, index_to_pack_map):
764
732
"""Determine how we want our packs to be ordered.
921
889
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
922
890
new_pack.signature_index.key_count(),
923
891
time.time() - new_pack.start_time)
925
# NB XXX: how to check CHK references are present? perhaps by yielding
926
# the items? How should that interact with stacked repos?
927
if new_pack.chk_index is not None:
929
if 'pack' in debug.debug_flags:
930
mutter('%s: create_pack: chk content copied: %s%s %d items t+%6.3fs',
931
time.ctime(), self._pack_collection._upload_transport.base,
932
new_pack.random_name,
933
new_pack.chk_index.key_count(),
934
time.time() - new_pack.start_time)
935
892
new_pack._check_references()
936
893
if not self._use_pack(new_pack):
941
898
self._pack_collection.allocate(new_pack)
944
def _copy_chks(self, refs=None):
945
# XXX: Todo, recursive follow-pointers facility when fetching some
947
chk_index_map, chk_indices = self._pack_map_and_index_list(
949
chk_nodes = self._index_contents(chk_indices, refs)
951
# TODO: This isn't strictly tasteful as we are accessing some private
952
# variables (_serializer). Perhaps a better way would be to have
953
# Repository._deserialise_chk_node()
954
search_key_func = chk_map.search_key_registry.get(
955
self._pack_collection.repo._serializer.search_key_name)
956
def accumlate_refs(lines):
957
# XXX: move to a generic location
959
bytes = ''.join(lines)
960
node = chk_map._deserialise(bytes, ("unknown",), search_key_func)
961
new_refs.update(node.refs())
962
self._copy_nodes(chk_nodes, chk_index_map, self.new_pack._writer,
963
self.new_pack.chk_index, output_lines=accumlate_refs)
966
def _copy_nodes(self, nodes, index_map, writer, write_index,
968
"""Copy knit nodes between packs with no graph references.
970
:param output_lines: Output full texts of copied items.
901
def _copy_nodes(self, nodes, index_map, writer, write_index):
902
"""Copy knit nodes between packs with no graph references."""
972
903
pb = ui.ui_factory.nested_progress_bar()
974
905
return self._do_copy_nodes(nodes, index_map, writer,
975
write_index, pb, output_lines=output_lines)
979
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb,
910
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
981
911
# for record verification
982
912
knit = KnitVersionedFiles(None, None)
983
913
# plan a readv on each source pack:
1383
1304
self._pack_transport = pack_transport
1384
1305
self._index_builder_class = index_builder_class
1385
1306
self._index_class = index_class
1386
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3,
1307
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1388
1308
self.packs = []
1389
1309
# name:Pack mapping
1391
1310
self._packs_by_name = {}
1392
1311
# the previous pack-names content
1393
1312
self._packs_at_load = None
1394
1313
# when a pack is being created by this object, the state of that pack.
1395
1314
self._new_pack = None
1396
1315
# 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]))
1316
self.revision_index = AggregateIndex(self.reload_pack_names)
1317
self.inventory_index = AggregateIndex(self.reload_pack_names)
1318
self.text_index = AggregateIndex(self.reload_pack_names)
1319
self.signature_index = AggregateIndex(self.reload_pack_names)
1416
1320
# resumed packs
1417
1321
self._resumed_packs = []
1420
return '%s(%r)' % (self.__class__.__name__, self.repo)
1422
1323
def add_pack_to_memory(self, pack):
1423
1324
"""Make a Pack object available to the repository to satisfy queries.
1558
1452
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
1455
"""Pack the pack collection totally."""
1566
1456
self.ensure_loaded()
1567
1457
total_packs = len(self._names)
1568
if self._already_packed():
1459
# This is arguably wrong because we might not be optimal, but for
1460
# now lets leave it in. (e.g. reconcile -> one pack. But not
1570
1463
total_revisions = self.revision_index.combined_index.key_count()
1571
1464
# XXX: the following may want to be a class, to pack with a given
1573
1466
mutter('Packing repository %s, which has %d pack files, '
1574
'containing %d revisions with hint %r.', self, total_packs,
1575
total_revisions, hint)
1467
'containing %d revisions into 1 packs.', self, total_packs,
1576
1469
# determine which packs need changing
1470
pack_distribution = [1]
1577
1471
pack_operations = [[0, []]]
1578
1472
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)
1473
pack_operations[-1][0] += pack.get_revision_count()
1474
pack_operations[-1][1].append(pack)
1584
1475
self._execute_pack_operations(pack_operations, OptimisingPacker)
1586
if clean_obsolete_packs:
1587
self._clear_obsolete_packs()
1589
1477
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1590
1478
"""Plan a pack operation.
1703
1579
inv_index = self._make_index(name, '.iix', resume=True)
1704
1580
txt_index = self._make_index(name, '.tix', resume=True)
1705
1581
sig_index = self._make_index(name, '.six', resume=True)
1706
if self.chk_index is not None:
1707
chk_index = self._make_index(name, '.cix', resume=True,
1708
unlimited_cache=True)
1711
result = self.resumed_pack_factory(name, rev_index, inv_index,
1712
txt_index, sig_index, self._upload_transport,
1713
self._pack_transport, self._index_transport, self,
1714
chk_index=chk_index)
1582
result = ResumedPack(name, rev_index, inv_index, txt_index,
1583
sig_index, self._upload_transport, self._pack_transport,
1584
self._index_transport, self)
1715
1585
except errors.NoSuchFile, e:
1716
1586
raise errors.UnresumableWriteGroup(self.repo, [name], str(e))
1717
1587
self.add_pack_to_memory(result)
1785
1654
:param return: None.
1787
1656
for pack in packs:
1789
pack.pack_transport.rename(pack.file_name(),
1790
'../obsolete_packs/' + pack.file_name())
1791
except (errors.PathError, errors.TransportError), e:
1792
# TODO: Should these be warnings or mutters?
1793
mutter("couldn't rename obsolete pack, skipping it:\n%s"
1657
pack.pack_transport.rename(pack.file_name(),
1658
'../obsolete_packs/' + pack.file_name())
1795
1659
# TODO: Probably needs to know all possible indices for this pack
1796
1660
# - or maybe list the directory and move all indices matching this
1797
1661
# name whether we recognize it or not?
1798
suffixes = ['.iix', '.six', '.tix', '.rix']
1799
if self.chk_index is not None:
1800
suffixes.append('.cix')
1801
for suffix in suffixes:
1803
self._index_transport.rename(pack.name + suffix,
1804
'../obsolete_packs/' + pack.name + suffix)
1805
except (errors.PathError, errors.TransportError), e:
1806
mutter("couldn't rename obsolete index, skipping it:\n%s"
1662
for suffix in ('.iix', '.six', '.tix', '.rix'):
1663
self._index_transport.rename(pack.name + suffix,
1664
'../obsolete_packs/' + pack.name + suffix)
1809
1666
def pack_distribution(self, total_revisions):
1810
1667
"""Generate a list of the number of revisions to put in each pack.
1836
1693
self._remove_pack_indices(pack)
1837
1694
self.packs.remove(pack)
1839
def _remove_pack_indices(self, pack, ignore_missing=False):
1840
"""Remove the indices for pack from the aggregated indices.
1842
:param ignore_missing: Suppress KeyErrors from calling remove_index.
1844
for index_type in Pack.index_definitions.keys():
1845
attr_name = index_type + '_index'
1846
aggregate_index = getattr(self, attr_name)
1847
if aggregate_index is not None:
1848
pack_index = getattr(pack, attr_name)
1850
aggregate_index.remove_index(pack_index)
1696
def _remove_pack_indices(self, pack):
1697
"""Remove the indices for pack from the aggregated indices."""
1698
self.revision_index.remove_index(pack.revision_index, pack)
1699
self.inventory_index.remove_index(pack.inventory_index, pack)
1700
self.text_index.remove_index(pack.text_index, pack)
1701
self.signature_index.remove_index(pack.signature_index, pack)
1856
1703
def reset(self):
1857
1704
"""Clear all cached data."""
1858
1705
# cached revision data
1706
self.repo._revision_knit = None
1859
1707
self.revision_index.clear()
1860
1708
# cached signature data
1709
self.repo._signature_knit = None
1861
1710
self.signature_index.clear()
1862
1711
# cached file text data
1863
1712
self.text_index.clear()
1713
self.repo._text_knit = None
1864
1714
# cached inventory data
1865
1715
self.inventory_index.clear()
1867
if self.chk_index is not None:
1868
self.chk_index.clear()
1869
1716
# remove the open pack
1870
1717
self._new_pack = None
1871
1718
# information about packs.
1982
1823
builder.add_node(key, value)
1983
1824
self.transport.put_file('pack-names', builder.finish(),
1984
1825
mode=self.repo.bzrdir._get_file_mode())
1826
# move the baseline forward
1985
1827
self._packs_at_load = disk_nodes
1986
1828
if clear_obsolete_packs:
1989
to_preserve = set([o.name for o in obsolete_packs])
1990
already_obsolete = self._clear_obsolete_packs(to_preserve)
1829
self._clear_obsolete_packs()
1992
1831
self._unlock_names()
1993
1832
# synchronise the memory packs list with what we just wrote:
1994
1833
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
1835
def reload_pack_names(self):
2007
1836
"""Sync our pack listing with what is present in the repository.
2009
1838
This should be called when we find out that something we thought was
2010
1839
present is now missing. This happens when another process re-packs the
2011
1840
repository, etc.
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()
1842
# This is functionally similar to _save_pack_names, but we don't write
2023
1843
# 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
1844
disk_nodes, _, _ = self._diff_pack_names()
1845
self._packs_at_load = disk_nodes
2030
1846
(removed, added,
2031
1847
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2032
1848
if removed or added or modified:
2042
1858
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2044
def _clear_obsolete_packs(self, preserve=None):
1860
def _clear_obsolete_packs(self):
2045
1861
"""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
1863
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2052
if preserve is None:
2054
1864
for filename in obsolete_pack_transport.list_dir('.'):
2055
name, ext = osutils.splitext(filename)
2058
if name in preserve:
2061
1866
obsolete_pack_transport.delete(filename)
2062
1867
except (errors.PathError, errors.TransportError), e:
2063
warning("couldn't delete obsolete pack, skipping it:\n%s"
1868
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2067
1870
def _start_write_group(self):
2068
1871
# Do not permit preparation for writing if we're not in a 'write lock'.
2069
1872
if not self.repo.is_write_locked():
2070
1873
raise errors.NotWriteLocked(self)
2071
self._new_pack = self.pack_factory(self, upload_suffix='.pack',
1874
self._new_pack = NewPack(self, upload_suffix='.pack',
2072
1875
file_mode=self.repo.bzrdir._get_file_mode())
2073
1876
# allow writing: queue writes to a new index
2074
1877
self.revision_index.add_writable_index(self._new_pack.revision_index,
2095
1892
# FIXME: just drop the transient index.
2096
1893
# forget what names there are
2097
1894
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()
1896
self._new_pack.abort()
1898
# XXX: If we aborted while in the middle of finishing the write
1899
# group, _remove_pack_indices can fail because the indexes are
1900
# already gone. If they're not there we shouldn't fail in this
1901
# case. -- mbp 20081113
1902
self._remove_pack_indices(self._new_pack)
1903
self._new_pack = None
2107
1904
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()
1906
resumed_pack.abort()
1908
# See comment in previous finally block.
1910
self._remove_pack_indices(resumed_pack)
2113
1913
del self._resumed_packs[:]
1914
self.repo._text_knit = None
2115
1916
def _remove_resumed_pack_indices(self):
2116
1917
for resumed_pack in self._resumed_packs:
2117
1918
self._remove_pack_indices(resumed_pack)
2118
1919
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
2130
1921
def _commit_write_group(self):
2131
1922
all_missing = set()
2132
1923
for prefix, versioned_file in (
2257
2039
deltas=True, parents=True, is_locked=self.is_locked),
2258
2040
data_access=self._pack_collection.text_index.data_access,
2259
2041
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
2272
2042
# True when the repository object is 'write locked' (as opposed to the
2273
2043
# physical lock only taken out around changes to the pack-names list.)
2274
2044
# Another way to represent this would be a decorator around the control
2281
2051
self._reconcile_fixes_text_parents = True
2282
2052
self._reconcile_backsup_inventory = False
2284
def _warn_if_deprecated(self, branch=None):
2054
def _warn_if_deprecated(self):
2285
2055
# This class isn't deprecated, but one sub-format is
2286
2056
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2287
super(KnitPackRepository, self)._warn_if_deprecated(branch)
2057
from bzrlib import repository
2058
if repository._deprecation_warning_done:
2060
repository._deprecation_warning_done = True
2061
warning("Format %s for %s is deprecated - please use"
2062
" 'bzr upgrade --1.6.1-rich-root'"
2063
% (self._format, self.bzrdir.transport.base))
2289
2065
def _abort_write_group(self):
2290
self.revisions._index._key_dependencies.clear()
2291
2066
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)
2068
def _find_inconsistent_revision_parents(self):
2069
"""Find revisions with incorrectly cached parents.
2071
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
2072
parents-in-revision).
2074
if not self.is_locked():
2075
raise errors.ObjectNotLocked(self)
2076
pb = ui.ui_factory.nested_progress_bar()
2079
revision_nodes = self._pack_collection.revision_index \
2080
.combined_index.iter_all_entries()
2081
index_positions = []
2082
# Get the cached index values for all revisions, and also the location
2083
# in each index of the revision text so we can perform linear IO.
2084
for index, key, value, refs in revision_nodes:
2085
pos, length = value[1:].split(' ')
2086
index_positions.append((index, int(pos), key[0],
2087
tuple(parent[0] for parent in refs[0])))
2088
pb.update("Reading revision index.", 0, 0)
2089
index_positions.sort()
2090
batch_count = len(index_positions) / 1000 + 1
2091
pb.update("Checking cached revision graph.", 0, batch_count)
2092
for offset in xrange(batch_count):
2093
pb.update("Checking cached revision graph.", offset)
2094
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
2097
rev_ids = [item[2] for item in to_query]
2098
revs = self.get_revisions(rev_ids)
2099
for revision, item in zip(revs, to_query):
2100
index_parents = item[3]
2101
rev_parents = tuple(revision.parent_ids)
2102
if index_parents != rev_parents:
2103
result.append((revision.revision_id, index_parents, rev_parents))
2108
@symbol_versioning.deprecated_method(symbol_versioning.one_one)
2109
def get_parents(self, revision_ids):
2110
"""See graph._StackedParentsProvider.get_parents."""
2111
parent_map = self.get_parent_map(revision_ids)
2112
return [parent_map.get(r, None) for r in revision_ids]
2298
2114
def _make_parents_provider(self):
2299
2115
return graph.CachingParentsProvider(self)
2301
2117
def _refresh_data(self):
2302
if not self.is_locked():
2304
self._pack_collection.reload_pack_names()
2118
if self._write_lock_count == 1 or (
2119
self.control_files._lock_count == 1 and
2120
self.control_files._lock_mode == 'r'):
2121
# forget what names there are
2122
self._pack_collection.reset()
2123
# XXX: Better to do an in-memory merge when acquiring a new lock -
2124
# factor out code from _save_pack_names.
2125
self._pack_collection.ensure_loaded()
2306
2127
def _start_write_group(self):
2307
2128
self._pack_collection._start_write_group()
2309
2130
def _commit_write_group(self):
2310
hint = self._pack_collection._commit_write_group()
2311
self.revisions._index._key_dependencies.clear()
2131
return self._pack_collection._commit_write_group()
2314
2133
def suspend_write_group(self):
2315
2134
# XXX check self._write_group is self.get_transaction()?
2316
2135
tokens = self._pack_collection._suspend_write_group()
2317
self.revisions._index._key_dependencies.clear()
2318
2136
self._write_group = None
2321
2139
def _resume_write_group(self, tokens):
2322
2140
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)
2141
self._pack_collection._resume_write_group(tokens)
2331
2143
def get_transaction(self):
2332
2144
if self._write_lock_count:
2341
2153
return self._write_lock_count
2343
2155
def lock_write(self, token=None):
2344
locked = self.is_locked()
2345
if not self._write_lock_count and locked:
2156
if not self._write_lock_count and self.is_locked():
2346
2157
raise errors.ReadOnlyError(self)
2347
2158
self._write_lock_count += 1
2348
2159
if self._write_lock_count == 1:
2349
2160
self._transaction = transactions.WriteTransaction()
2351
if 'relock' in debug.debug_flags and self._prev_lock == 'w':
2352
note('%r was write locked again', self)
2353
self._prev_lock = 'w'
2354
2161
for repo in self._fallback_repositories:
2355
2162
# Writes don't affect fallback repos
2356
2163
repo.lock_read()
2357
self._refresh_data()
2358
return RepositoryWriteLockResult(self.unlock, None)
2164
self._refresh_data()
2360
2166
def lock_read(self):
2361
locked = self.is_locked()
2362
2167
if self._write_lock_count:
2363
2168
self._write_lock_count += 1
2365
2170
self.control_files.lock_read()
2367
if 'relock' in debug.debug_flags and self._prev_lock == 'r':
2368
note('%r was read locked again', self)
2369
self._prev_lock = 'r'
2370
2171
for repo in self._fallback_repositories:
2172
# Writes don't affect fallback repos
2371
2173
repo.lock_read()
2372
self._refresh_data()
2174
self._refresh_data()
2375
2176
def leave_lock_in_place(self):
2376
2177
# not supported - raise an error
2416
2212
transaction = self._transaction
2417
2213
self._transaction = None
2418
2214
transaction.finish()
2215
for repo in self._fallback_repositories:
2420
2218
self.control_files.unlock()
2422
if not self.is_locked():
2423
2219
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
2223
class RepositoryFormatPack(MetaDirRepositoryFormat):
2501
2224
"""Format logic for pack structured repositories.
2880
2632
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2635
class RepositoryFormatPackDevelopment2(RepositoryFormatPack):
2636
"""A no-subtrees development repository.
2638
This format should be retained until the second release after bzr 1.7.
2640
This is pack-1.6.1 with B+Tree indices.
2643
repository_class = KnitPackRepository
2644
_commit_builder_class = PackCommitBuilder
2645
supports_external_lookups = True
2646
# What index classes to use
2647
index_builder_class = BTreeBuilder
2648
index_class = BTreeGraphIndex
2651
def _serializer(self):
2652
return xml5.serializer_v5
2654
def _get_matching_bzrdir(self):
2655
return bzrdir.format_registry.make_bzrdir('development2')
2657
def _ignore_setting_bzrdir(self, format):
2660
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2662
def get_format_string(self):
2663
"""See RepositoryFormat.get_format_string()."""
2664
return "Bazaar development format 2 (needs bzr.dev from before 1.8)\n"
2666
def get_format_description(self):
2667
"""See RepositoryFormat.get_format_description()."""
2668
return ("Development repository format, currently the same as "
2669
"1.6.1 with B+Trees.\n")
2671
def check_conversion_target(self, target_format):
2883
2675
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2884
2676
"""A subtrees development repository.
2886
2678
This format should be retained until the second release after bzr 1.7.
2888
2680
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
2683
repository_class = KnitPackRepository
2895
2684
_commit_builder_class = PackRootCommitBuilder
2896
2685
rich_root_data = True
2898
2686
supports_tree_reference = True
2899
2687
supports_external_lookups = True
2900
2688
# What index classes to use
2908
2696
def _get_matching_bzrdir(self):
2909
2697
return bzrdir.format_registry.make_bzrdir(
2910
'development-subtree')
2698
'development2-subtree')
2912
2700
def _ignore_setting_bzrdir(self, format):
2915
2703
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2705
def check_conversion_target(self, target_format):
2706
if not target_format.rich_root_data:
2707
raise errors.BadConversionTarget(
2708
'Does not support rich root data.', target_format)
2709
if not getattr(target_format, 'supports_tree_reference', False):
2710
raise errors.BadConversionTarget(
2711
'Does not support nested trees', target_format)
2917
2713
def get_format_string(self):
2918
2714
"""See RepositoryFormat.get_format_string()."""
2919
2715
return ("Bazaar development format 2 with subtree support "
added
removed
bzrlib/repofmt/pack_repo.py
