152
151
texts/deltas (via (fileid, revisionid) tuples).
153
152
:param signature_index: A GraphIndex for determining what signatures are
154
153
present in the Pack and accessing the locations of their texts.
155
:param chk_index: A GraphIndex for accessing content by CHK, if the
158
155
self.revision_index = revision_index
159
156
self.inventory_index = inventory_index
160
157
self.text_index = text_index
161
158
self.signature_index = signature_index
162
self.chk_index = chk_index
164
160
def access_tuple(self):
165
161
"""Return a tuple (transport, name) for the pack content."""
226
222
return self.index_name('text', name)
228
224
def _replace_index_with_readonly(self, index_type):
229
unlimited_cache = False
230
if index_type == 'chk':
231
unlimited_cache = True
232
225
setattr(self, index_type + '_index',
233
226
self.index_class(self.index_transport,
234
227
self.index_name(index_type, self.name),
235
self.index_sizes[self.index_offset(index_type)],
236
unlimited_cache=unlimited_cache))
228
self.index_sizes[self.index_offset(index_type)]))
239
231
class ExistingPack(Pack):
240
232
"""An in memory proxy for an existing .pack and its disk indices."""
242
234
def __init__(self, pack_transport, name, revision_index, inventory_index,
243
text_index, signature_index, chk_index=None):
235
text_index, signature_index):
244
236
"""Create an ExistingPack object.
246
238
:param pack_transport: The transport where the pack file resides.
247
239
:param name: The name of the pack on disk in the pack_transport.
249
241
Pack.__init__(self, revision_index, inventory_index, text_index,
250
signature_index, chk_index)
252
244
self.pack_transport = pack_transport
253
245
if None in (revision_index, inventory_index, text_index,
271
263
def __init__(self, name, revision_index, inventory_index, text_index,
272
264
signature_index, upload_transport, pack_transport, index_transport,
273
pack_collection, chk_index=None):
274
266
"""Create a ResumedPack object."""
275
267
ExistingPack.__init__(self, pack_transport, name, revision_index,
276
inventory_index, text_index, signature_index,
268
inventory_index, text_index, signature_index)
278
269
self.upload_transport = upload_transport
279
270
self.index_transport = index_transport
280
271
self.index_sizes = [None, None, None, None]
307
295
self.upload_transport.delete(self.file_name())
308
296
indices = [self.revision_index, self.inventory_index, self.text_index,
309
297
self.signature_index]
310
if self.chk_index is not None:
311
indices.append(self.chk_index)
312
298
for index in indices:
313
299
index._transport.delete(index._name)
315
301
def finish(self):
316
302
self._check_references()
317
index_types = ['revision', 'inventory', 'text', 'signature']
318
if self.chk_index is not None:
319
index_types.append('chk')
320
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']:
321
306
old_name = self.index_name(index_type, self.name)
322
307
new_name = '../indices/' + old_name
323
308
self.upload_transport.rename(old_name, new_name)
324
309
self._replace_index_with_readonly(index_type)
325
new_name = '../packs/' + self.file_name()
326
self.upload_transport.rename(self.file_name(), new_name)
327
310
self._state = 'finished'
329
312
def _get_external_refs(self, index):
330
"""Return compression parents for this index that are not present.
332
This returns any compression parents that are referenced by this index,
333
which are not contained *in* this index. They may be present elsewhere.
335
313
return index.external_references(1)
383
355
self._file_mode = file_mode
384
356
# tracks the content written to the .pack file.
385
357
self._hash = osutils.md5()
386
# a tuple with the length in bytes of the indices, once the pack
387
# is finalised. (rev, inv, text, sigs, chk_if_in_use)
358
# a four-tuple with the length in bytes of the indices, once the pack
359
# is finalised. (rev, inv, text, sigs)
388
360
self.index_sizes = None
389
361
# How much data to cache when writing packs. Note that this is not
390
362
# synchronised with reads, because it's not in the transport layer, so
453
423
return bool(self.get_revision_count() or
454
424
self.inventory_index.key_count() or
455
425
self.text_index.key_count() or
456
self.signature_index.key_count() or
457
(self.chk_index is not None and self.chk_index.key_count()))
459
def finish_content(self):
460
if self.name is not None:
464
self._write_data('', flush=True)
465
self.name = self._hash.hexdigest()
426
self.signature_index.key_count())
467
428
def finish(self, suspend=False):
468
429
"""Finish the new pack.
490
454
self._write_index('text', self.text_index, 'file texts', suspend)
491
455
self._write_index('signature', self.signature_index,
492
456
'revision signatures', suspend)
493
if self.chk_index is not None:
494
self.index_sizes.append(None)
495
self._write_index('chk', self.chk_index,
496
'content hash bytes', suspend)
497
457
self.write_stream.close()
498
458
# Note that this will clobber an existing pack with the same name,
499
459
# without checking for hash collisions. While this is undesirable this
583
543
self.index_to_pack = {}
584
544
self.combined_index = CombinedGraphIndex([], reload_func=reload_func)
585
545
self.data_access = _DirectPackAccess(self.index_to_pack,
586
reload_func=reload_func,
587
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
588
567
self.add_callback = None
590
569
def add_index(self, index, pack):
625
604
self.data_access.set_writer(None, None, (None, None))
626
605
self.index_to_pack.clear()
627
606
del self.combined_index._indices[:]
628
del self.combined_index._index_names[:]
629
607
self.add_callback = None
631
def remove_index(self, index):
609
def remove_index(self, index, pack):
632
610
"""Remove index from the indices used to answer queries.
634
612
:param index: An index from the pack parameter.
613
:param pack: A Pack instance.
636
615
del self.index_to_pack[index]
637
pos = self.combined_index._indices.index(index)
638
del self.combined_index._indices[pos]
639
del self.combined_index._index_names[pos]
616
self.combined_index._indices.remove(index)
640
617
if (self.add_callback is not None and
641
618
getattr(index, 'add_nodes', None) == self.add_callback):
642
619
self.add_callback = None
920
896
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
921
897
new_pack.signature_index.key_count(),
922
898
time.time() - new_pack.start_time)
924
# NB XXX: how to check CHK references are present? perhaps by yielding
925
# the items? How should that interact with stacked repos?
926
if new_pack.chk_index is not None:
928
if 'pack' in debug.debug_flags:
929
mutter('%s: create_pack: chk content copied: %s%s %d items t+%6.3fs',
930
time.ctime(), self._pack_collection._upload_transport.base,
931
new_pack.random_name,
932
new_pack.chk_index.key_count(),
933
time.time() - new_pack.start_time)
934
899
new_pack._check_references()
935
900
if not self._use_pack(new_pack):
940
905
self._pack_collection.allocate(new_pack)
943
def _copy_chks(self, refs=None):
944
# XXX: Todo, recursive follow-pointers facility when fetching some
946
chk_index_map, chk_indices = self._pack_map_and_index_list(
948
chk_nodes = self._index_contents(chk_indices, refs)
950
# TODO: This isn't strictly tasteful as we are accessing some private
951
# variables (_serializer). Perhaps a better way would be to have
952
# Repository._deserialise_chk_node()
953
search_key_func = chk_map.search_key_registry.get(
954
self._pack_collection.repo._serializer.search_key_name)
955
def accumlate_refs(lines):
956
# XXX: move to a generic location
958
bytes = ''.join(lines)
959
node = chk_map._deserialise(bytes, ("unknown",), search_key_func)
960
new_refs.update(node.refs())
961
self._copy_nodes(chk_nodes, chk_index_map, self.new_pack._writer,
962
self.new_pack.chk_index, output_lines=accumlate_refs)
965
def _copy_nodes(self, nodes, index_map, writer, write_index,
967
"""Copy knit nodes between packs with no graph references.
969
:param output_lines: Output full texts of copied items.
908
def _copy_nodes(self, nodes, index_map, writer, write_index):
909
"""Copy knit nodes between packs with no graph references."""
971
910
pb = ui.ui_factory.nested_progress_bar()
973
912
return self._do_copy_nodes(nodes, index_map, writer,
974
write_index, pb, output_lines=output_lines)
978
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb,
917
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
980
918
# for record verification
981
919
knit = KnitVersionedFiles(None, None)
982
920
# plan a readv on each source pack:
1016
954
izip(reader.iter_records(), pack_readv_requests):
1017
955
raw_data = read_func(None)
1018
956
# check the header only
1019
if output_lines is not None:
1020
output_lines(knit._parse_record(key[-1], raw_data)[0])
1022
df, _ = knit._parse_record_header(key, raw_data)
957
df, _ = knit._parse_record_header(key, raw_data)
1024
959
pos, size = writer.add_bytes_record(raw_data, names)
1025
960
write_index.add_node(key, eol_flag + "%d %d" % (pos, size))
1026
961
pb.update("Copied record", record_index)
1356
1291
:ivar _names: map of {pack_name: (index_size,)}
1359
pack_factory = NewPack
1360
resumed_pack_factory = ResumedPack
1362
1294
def __init__(self, repo, transport, index_transport, upload_transport,
1363
pack_transport, index_builder_class, index_class,
1295
pack_transport, index_builder_class, index_class):
1365
1296
"""Create a new RepositoryPackCollection.
1367
1298
:param transport: Addresses the repository base directory
1393
1322
# when a pack is being created by this object, the state of that pack.
1394
1323
self._new_pack = None
1395
1324
# aggregated revision index data
1396
flush = self._flush_new_pack
1397
self.revision_index = AggregateIndex(self.reload_pack_names, flush)
1398
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1399
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1400
self.signature_index = AggregateIndex(self.reload_pack_names, flush)
1401
all_indices = [self.revision_index, self.inventory_index,
1402
self.text_index, self.signature_index]
1404
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1405
all_indices.append(self.chk_index)
1407
# used to determine if we're using a chk_index elsewhere.
1408
self.chk_index = None
1409
# Tell all the CombinedGraphIndex objects about each other, so they can
1410
# share hints about which pack names to search first.
1411
all_combined = [agg_idx.combined_index for agg_idx in all_indices]
1412
for combined_idx in all_combined:
1413
combined_idx.set_sibling_indices(
1414
set(all_combined).difference([combined_idx]))
1325
self.revision_index = AggregateIndex(self.reload_pack_names)
1326
self.inventory_index = AggregateIndex(self.reload_pack_names)
1327
self.text_index = AggregateIndex(self.reload_pack_names)
1328
self.signature_index = AggregateIndex(self.reload_pack_names)
1415
1329
# resumed packs
1416
1330
self._resumed_packs = []
1419
return '%s(%r)' % (self.__class__.__name__, self.repo)
1421
1332
def add_pack_to_memory(self, pack):
1422
1333
"""Make a Pack object available to the repository to satisfy queries.
1504
1415
'containing %d revisions. Packing %d files into %d affecting %d'
1505
1416
' revisions', self, total_packs, total_revisions, num_old_packs,
1506
1417
num_new_packs, num_revs_affected)
1507
result = self._execute_pack_operations(pack_operations,
1418
self._execute_pack_operations(pack_operations,
1508
1419
reload_func=self._restart_autopack)
1509
mutter('Auto-packing repository %s completed', self)
1512
1422
def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1513
1423
reload_func=None):
1537
1447
self._remove_pack_from_memory(pack)
1538
1448
# record the newly available packs and stop advertising the old
1540
to_be_obsoleted = []
1541
for _, packs in pack_operations:
1542
to_be_obsoleted.extend(packs)
1543
result = self._save_pack_names(clear_obsolete_packs=True,
1544
obsolete_packs=to_be_obsoleted)
1547
def _flush_new_pack(self):
1548
if self._new_pack is not None:
1549
self._new_pack.flush()
1450
self._save_pack_names(clear_obsolete_packs=True)
1451
# Move the old packs out of the way now they are no longer referenced.
1452
for revision_count, packs in pack_operations:
1453
self._obsolete_packs(packs)
1551
1455
def lock_names(self):
1552
1456
"""Acquire the mutex around the pack-names index.
1557
1461
self.repo.control_files.lock_write()
1559
def _already_packed(self):
1560
"""Is the collection already packed?"""
1561
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1563
def pack(self, hint=None, clean_obsolete_packs=False):
1564
1464
"""Pack the pack collection totally."""
1565
1465
self.ensure_loaded()
1566
1466
total_packs = len(self._names)
1567
if self._already_packed():
1468
# This is arguably wrong because we might not be optimal, but for
1469
# now lets leave it in. (e.g. reconcile -> one pack. But not
1569
1472
total_revisions = self.revision_index.combined_index.key_count()
1570
1473
# XXX: the following may want to be a class, to pack with a given
1572
1475
mutter('Packing repository %s, which has %d pack files, '
1573
'containing %d revisions with hint %r.', self, total_packs,
1574
total_revisions, hint)
1476
'containing %d revisions into 1 packs.', self, total_packs,
1575
1478
# determine which packs need changing
1479
pack_distribution = [1]
1576
1480
pack_operations = [[0, []]]
1577
1481
for pack in self.all_packs():
1578
if hint is None or pack.name in hint:
1579
# Either no hint was provided (so we are packing everything),
1580
# or this pack was included in the hint.
1581
pack_operations[-1][0] += pack.get_revision_count()
1582
pack_operations[-1][1].append(pack)
1482
pack_operations[-1][0] += pack.get_revision_count()
1483
pack_operations[-1][1].append(pack)
1583
1484
self._execute_pack_operations(pack_operations, OptimisingPacker)
1585
if clean_obsolete_packs:
1586
self._clear_obsolete_packs()
1588
1486
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1589
1487
"""Plan a pack operation.
1677
1575
inv_index = self._make_index(name, '.iix')
1678
1576
txt_index = self._make_index(name, '.tix')
1679
1577
sig_index = self._make_index(name, '.six')
1680
if self.chk_index is not None:
1681
chk_index = self._make_index(name, '.cix', unlimited_cache=True)
1684
1578
result = ExistingPack(self._pack_transport, name, rev_index,
1685
inv_index, txt_index, sig_index, chk_index)
1579
inv_index, txt_index, sig_index)
1686
1580
self.add_pack_to_memory(result)
1702
1596
inv_index = self._make_index(name, '.iix', resume=True)
1703
1597
txt_index = self._make_index(name, '.tix', resume=True)
1704
1598
sig_index = self._make_index(name, '.six', resume=True)
1705
if self.chk_index is not None:
1706
chk_index = self._make_index(name, '.cix', resume=True,
1707
unlimited_cache=True)
1710
result = self.resumed_pack_factory(name, rev_index, inv_index,
1711
txt_index, sig_index, self._upload_transport,
1712
self._pack_transport, self._index_transport, self,
1713
chk_index=chk_index)
1599
result = ResumedPack(name, rev_index, inv_index, txt_index,
1600
sig_index, self._upload_transport, self._pack_transport,
1601
self._index_transport, self)
1714
1602
except errors.NoSuchFile, e:
1715
1603
raise errors.UnresumableWriteGroup(self.repo, [name], str(e))
1716
1604
self.add_pack_to_memory(result)
1784
1671
:param return: None.
1786
1673
for pack in packs:
1788
pack.pack_transport.rename(pack.file_name(),
1789
'../obsolete_packs/' + pack.file_name())
1790
except (errors.PathError, errors.TransportError), e:
1791
# TODO: Should these be warnings or mutters?
1792
mutter("couldn't rename obsolete pack, skipping it:\n%s"
1674
pack.pack_transport.rename(pack.file_name(),
1675
'../obsolete_packs/' + pack.file_name())
1794
1676
# TODO: Probably needs to know all possible indices for this pack
1795
1677
# - or maybe list the directory and move all indices matching this
1796
1678
# name whether we recognize it or not?
1797
suffixes = ['.iix', '.six', '.tix', '.rix']
1798
if self.chk_index is not None:
1799
suffixes.append('.cix')
1800
for suffix in suffixes:
1802
self._index_transport.rename(pack.name + suffix,
1803
'../obsolete_packs/' + pack.name + suffix)
1804
except (errors.PathError, errors.TransportError), e:
1805
mutter("couldn't rename obsolete index, skipping it:\n%s"
1679
for suffix in ('.iix', '.six', '.tix', '.rix'):
1680
self._index_transport.rename(pack.name + suffix,
1681
'../obsolete_packs/' + pack.name + suffix)
1808
1683
def pack_distribution(self, total_revisions):
1809
1684
"""Generate a list of the number of revisions to put in each pack.
1835
1710
self._remove_pack_indices(pack)
1836
1711
self.packs.remove(pack)
1838
def _remove_pack_indices(self, pack, ignore_missing=False):
1839
"""Remove the indices for pack from the aggregated indices.
1841
:param ignore_missing: Suppress KeyErrors from calling remove_index.
1843
for index_type in Pack.index_definitions.keys():
1844
attr_name = index_type + '_index'
1845
aggregate_index = getattr(self, attr_name)
1846
if aggregate_index is not None:
1847
pack_index = getattr(pack, attr_name)
1849
aggregate_index.remove_index(pack_index)
1713
def _remove_pack_indices(self, pack):
1714
"""Remove the indices for pack from the aggregated indices."""
1715
self.revision_index.remove_index(pack.revision_index, pack)
1716
self.inventory_index.remove_index(pack.inventory_index, pack)
1717
self.text_index.remove_index(pack.text_index, pack)
1718
self.signature_index.remove_index(pack.signature_index, pack)
1855
1720
def reset(self):
1856
1721
"""Clear all cached data."""
1857
1722
# cached revision data
1723
self.repo._revision_knit = None
1858
1724
self.revision_index.clear()
1859
1725
# cached signature data
1726
self.repo._signature_knit = None
1860
1727
self.signature_index.clear()
1861
1728
# cached file text data
1862
1729
self.text_index.clear()
1730
self.repo._text_knit = None
1863
1731
# cached inventory data
1864
1732
self.inventory_index.clear()
1866
if self.chk_index is not None:
1867
self.chk_index.clear()
1868
1733
# remove the open pack
1869
1734
self._new_pack = None
1870
1735
# information about packs.
1965
1829
:param clear_obsolete_packs: If True, clear out the contents of the
1966
1830
obsolete_packs directory.
1967
:param obsolete_packs: Packs that are obsolete once the new pack-names
1968
file has been written.
1969
:return: A list of the names saved that were not previously on disk.
1971
already_obsolete = []
1972
1832
self.lock_names()
1974
1834
builder = self._index_builder_class()
1975
(disk_nodes, deleted_nodes, new_nodes,
1976
orig_disk_nodes) = self._diff_pack_names()
1835
disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1977
1836
# TODO: handle same-name, index-size-changes here -
1978
1837
# e.g. use the value from disk, not ours, *unless* we're the one
1981
1840
builder.add_node(key, value)
1982
1841
self.transport.put_file('pack-names', builder.finish(),
1983
1842
mode=self.repo.bzrdir._get_file_mode())
1843
# move the baseline forward
1984
1844
self._packs_at_load = disk_nodes
1985
1845
if clear_obsolete_packs:
1988
to_preserve = set([o.name for o in obsolete_packs])
1989
already_obsolete = self._clear_obsolete_packs(to_preserve)
1846
self._clear_obsolete_packs()
1991
1848
self._unlock_names()
1992
1849
# synchronise the memory packs list with what we just wrote:
1993
1850
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1995
# TODO: We could add one more condition here. "if o.name not in
1996
# orig_disk_nodes and o != the new_pack we haven't written to
1997
# disk yet. However, the new pack object is not easily
1998
# accessible here (it would have to be passed through the
1999
# autopacking code, etc.)
2000
obsolete_packs = [o for o in obsolete_packs
2001
if o.name not in already_obsolete]
2002
self._obsolete_packs(obsolete_packs)
2003
return [new_node[0][0] for new_node in new_nodes]
2005
1852
def reload_pack_names(self):
2006
1853
"""Sync our pack listing with what is present in the repository.
2022
1869
# out the new value.
2023
(disk_nodes, deleted_nodes, new_nodes,
2024
orig_disk_nodes) = self._diff_pack_names()
2025
# _packs_at_load is meant to be the explicit list of names in
2026
# 'pack-names' at then start. As such, it should not contain any
2027
# pending names that haven't been written out yet.
2028
self._packs_at_load = orig_disk_nodes
1870
disk_nodes, _, _ = self._diff_pack_names()
1871
self._packs_at_load = disk_nodes
2029
1872
(removed, added,
2030
1873
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2031
1874
if removed or added or modified:
2041
1884
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2043
def _clear_obsolete_packs(self, preserve=None):
1886
def _clear_obsolete_packs(self):
2044
1887
"""Delete everything from the obsolete-packs directory.
2046
:return: A list of pack identifiers (the filename without '.pack') that
2047
were found in obsolete_packs.
2050
1889
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2051
if preserve is None:
2053
1890
for filename in obsolete_pack_transport.list_dir('.'):
2054
name, ext = osutils.splitext(filename)
2057
if name in preserve:
2060
1892
obsolete_pack_transport.delete(filename)
2061
1893
except (errors.PathError, errors.TransportError), e:
2062
warning("couldn't delete obsolete pack, skipping it:\n%s"
1894
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2066
1896
def _start_write_group(self):
2067
1897
# Do not permit preparation for writing if we're not in a 'write lock'.
2068
1898
if not self.repo.is_write_locked():
2069
1899
raise errors.NotWriteLocked(self)
2070
self._new_pack = self.pack_factory(self, upload_suffix='.pack',
1900
self._new_pack = NewPack(self, upload_suffix='.pack',
2071
1901
file_mode=self.repo.bzrdir._get_file_mode())
2072
1902
# allow writing: queue writes to a new index
2073
1903
self.revision_index.add_writable_index(self._new_pack.revision_index,
2076
1906
self._new_pack)
2077
1907
self.text_index.add_writable_index(self._new_pack.text_index,
2078
1908
self._new_pack)
2079
self._new_pack.text_index.set_optimize(combine_backing_indices=False)
2080
1909
self.signature_index.add_writable_index(self._new_pack.signature_index,
2081
1910
self._new_pack)
2082
if self.chk_index is not None:
2083
self.chk_index.add_writable_index(self._new_pack.chk_index,
2085
self.repo.chk_bytes._index._add_callback = self.chk_index.add_callback
2086
self._new_pack.chk_index.set_optimize(combine_backing_indices=False)
2088
1912
self.repo.inventories._index._add_callback = self.inventory_index.add_callback
2089
1913
self.repo.revisions._index._add_callback = self.revision_index.add_callback
2094
1918
# FIXME: just drop the transient index.
2095
1919
# forget what names there are
2096
1920
if self._new_pack is not None:
2097
operation = cleanup.OperationWithCleanups(self._new_pack.abort)
2098
operation.add_cleanup(setattr, self, '_new_pack', None)
2099
# If we aborted while in the middle of finishing the write
2100
# group, _remove_pack_indices could fail because the indexes are
2101
# already gone. But they're not there we shouldn't fail in this
2102
# case, so we pass ignore_missing=True.
2103
operation.add_cleanup(self._remove_pack_indices, self._new_pack,
2104
ignore_missing=True)
2105
operation.run_simple()
1922
self._new_pack.abort()
1924
# XXX: If we aborted while in the middle of finishing the write
1925
# group, _remove_pack_indices can fail because the indexes are
1926
# already gone. If they're not there we shouldn't fail in this
1927
# case. -- mbp 20081113
1928
self._remove_pack_indices(self._new_pack)
1929
self._new_pack = None
2106
1930
for resumed_pack in self._resumed_packs:
2107
operation = cleanup.OperationWithCleanups(resumed_pack.abort)
2108
# See comment in previous finally block.
2109
operation.add_cleanup(self._remove_pack_indices, resumed_pack,
2110
ignore_missing=True)
2111
operation.run_simple()
1932
resumed_pack.abort()
1934
# See comment in previous finally block.
1936
self._remove_pack_indices(resumed_pack)
2112
1939
del self._resumed_packs[:]
1940
self.repo._text_knit = None
2114
1942
def _remove_resumed_pack_indices(self):
2115
1943
for resumed_pack in self._resumed_packs:
2116
1944
self._remove_pack_indices(resumed_pack)
2117
1945
del self._resumed_packs[:]
2119
def _check_new_inventories(self):
2120
"""Detect missing inventories in this write group.
2122
:returns: list of strs, summarising any problems found. If the list is
2123
empty no problems were found.
2125
# The base implementation does no checks. GCRepositoryPackCollection
2129
1947
def _commit_write_group(self):
2130
1948
all_missing = set()
2131
1949
for prefix, versioned_file in (
2140
1958
raise errors.BzrCheckError(
2141
1959
"Repository %s has missing compression parent(s) %r "
2142
1960
% (self.repo, sorted(all_missing)))
2143
problems = self._check_new_inventories()
2145
problems_summary = '\n'.join(problems)
2146
raise errors.BzrCheckError(
2147
"Cannot add revision(s) to repository: " + problems_summary)
2148
1961
self._remove_pack_indices(self._new_pack)
2149
any_new_content = False
1962
should_autopack = False
2150
1963
if self._new_pack.data_inserted():
2151
1964
# get all the data to disk and read to use
2152
1965
self._new_pack.finish()
2153
1966
self.allocate(self._new_pack)
2154
1967
self._new_pack = None
2155
any_new_content = True
1968
should_autopack = True
2157
1970
self._new_pack.abort()
2158
1971
self._new_pack = None
2240
2050
self.revisions = KnitVersionedFiles(
2241
2051
_KnitGraphIndex(self._pack_collection.revision_index.combined_index,
2242
2052
add_callback=self._pack_collection.revision_index.add_callback,
2243
deltas=False, parents=True, is_locked=self.is_locked,
2244
track_external_parent_refs=True),
2053
deltas=False, parents=True, is_locked=self.is_locked),
2245
2054
data_access=self._pack_collection.revision_index.data_access,
2246
2055
max_delta_chain=0)
2247
2056
self.signatures = KnitVersionedFiles(
2256
2065
deltas=True, parents=True, is_locked=self.is_locked),
2257
2066
data_access=self._pack_collection.text_index.data_access,
2258
2067
max_delta_chain=200)
2259
if _format.supports_chks:
2260
# No graph, no compression:- references from chks are between
2261
# different objects not temporal versions of the same; and without
2262
# some sort of temporal structure knit compression will just fail.
2263
self.chk_bytes = KnitVersionedFiles(
2264
_KnitGraphIndex(self._pack_collection.chk_index.combined_index,
2265
add_callback=self._pack_collection.chk_index.add_callback,
2266
deltas=False, parents=False, is_locked=self.is_locked),
2267
data_access=self._pack_collection.chk_index.data_access,
2270
self.chk_bytes = None
2271
2068
# True when the repository object is 'write locked' (as opposed to the
2272
2069
# physical lock only taken out around changes to the pack-names list.)
2273
2070
# Another way to represent this would be a decorator around the control
2280
2077
self._reconcile_fixes_text_parents = True
2281
2078
self._reconcile_backsup_inventory = False
2283
def _warn_if_deprecated(self, branch=None):
2080
def _warn_if_deprecated(self):
2284
2081
# This class isn't deprecated, but one sub-format is
2285
2082
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2286
super(KnitPackRepository, self)._warn_if_deprecated(branch)
2083
from bzrlib import repository
2084
if repository._deprecation_warning_done:
2086
repository._deprecation_warning_done = True
2087
warning("Format %s for %s is deprecated - please use"
2088
" 'bzr upgrade --1.6.1-rich-root'"
2089
% (self._format, self.bzrdir.transport.base))
2288
2091
def _abort_write_group(self):
2289
self.revisions._index._key_dependencies.clear()
2290
2092
self._pack_collection._abort_write_group()
2292
def _get_source(self, to_format):
2293
if to_format.network_name() == self._format.network_name():
2294
return KnitPackStreamSource(self, to_format)
2295
return super(KnitPackRepository, self)._get_source(to_format)
2094
def _find_inconsistent_revision_parents(self):
2095
"""Find revisions with incorrectly cached parents.
2097
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
2098
parents-in-revision).
2100
if not self.is_locked():
2101
raise errors.ObjectNotLocked(self)
2102
pb = ui.ui_factory.nested_progress_bar()
2105
revision_nodes = self._pack_collection.revision_index \
2106
.combined_index.iter_all_entries()
2107
index_positions = []
2108
# Get the cached index values for all revisions, and also the location
2109
# in each index of the revision text so we can perform linear IO.
2110
for index, key, value, refs in revision_nodes:
2111
pos, length = value[1:].split(' ')
2112
index_positions.append((index, int(pos), key[0],
2113
tuple(parent[0] for parent in refs[0])))
2114
pb.update("Reading revision index", 0, 0)
2115
index_positions.sort()
2116
batch_count = len(index_positions) / 1000 + 1
2117
pb.update("Checking cached revision graph", 0, batch_count)
2118
for offset in xrange(batch_count):
2119
pb.update("Checking cached revision graph", offset)
2120
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
2123
rev_ids = [item[2] for item in to_query]
2124
revs = self.get_revisions(rev_ids)
2125
for revision, item in zip(revs, to_query):
2126
index_parents = item[3]
2127
rev_parents = tuple(revision.parent_ids)
2128
if index_parents != rev_parents:
2129
result.append((revision.revision_id, index_parents, rev_parents))
2297
2134
def _make_parents_provider(self):
2298
2135
return graph.CachingParentsProvider(self)
2306
2143
self._pack_collection._start_write_group()
2308
2145
def _commit_write_group(self):
2309
hint = self._pack_collection._commit_write_group()
2310
self.revisions._index._key_dependencies.clear()
2146
return self._pack_collection._commit_write_group()
2313
2148
def suspend_write_group(self):
2314
2149
# XXX check self._write_group is self.get_transaction()?
2315
2150
tokens = self._pack_collection._suspend_write_group()
2316
self.revisions._index._key_dependencies.clear()
2317
2151
self._write_group = None
2320
2154
def _resume_write_group(self, tokens):
2321
2155
self._start_write_group()
2323
self._pack_collection._resume_write_group(tokens)
2324
except errors.UnresumableWriteGroup:
2325
self._abort_write_group()
2327
for pack in self._pack_collection._resumed_packs:
2328
self.revisions._index.scan_unvalidated_index(pack.revision_index)
2156
self._pack_collection._resume_write_group(tokens)
2330
2158
def get_transaction(self):
2331
2159
if self._write_lock_count:
2413
2231
transaction = self._transaction
2414
2232
self._transaction = None
2415
2233
transaction.finish()
2234
for repo in self._fallback_repositories:
2417
2237
self.control_files.unlock()
2419
if not self.is_locked():
2420
2238
for repo in self._fallback_repositories:
2424
class KnitPackStreamSource(StreamSource):
2425
"""A StreamSource used to transfer data between same-format KnitPack repos.
2427
This source assumes:
2428
1) Same serialization format for all objects
2429
2) Same root information
2430
3) XML format inventories
2431
4) Atomic inserts (so we can stream inventory texts before text
2436
def __init__(self, from_repository, to_format):
2437
super(KnitPackStreamSource, self).__init__(from_repository, to_format)
2438
self._text_keys = None
2439
self._text_fetch_order = 'unordered'
2441
def _get_filtered_inv_stream(self, revision_ids):
2442
from_repo = self.from_repository
2443
parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
2444
parent_keys = [(p,) for p in parent_ids]
2445
find_text_keys = from_repo._find_text_key_references_from_xml_inventory_lines
2446
parent_text_keys = set(find_text_keys(
2447
from_repo._inventory_xml_lines_for_keys(parent_keys)))
2448
content_text_keys = set()
2449
knit = KnitVersionedFiles(None, None)
2450
factory = KnitPlainFactory()
2451
def find_text_keys_from_content(record):
2452
if record.storage_kind not in ('knit-delta-gz', 'knit-ft-gz'):
2453
raise ValueError("Unknown content storage kind for"
2454
" inventory text: %s" % (record.storage_kind,))
2455
# It's a knit record, it has a _raw_record field (even if it was
2456
# reconstituted from a network stream).
2457
raw_data = record._raw_record
2458
# read the entire thing
2459
revision_id = record.key[-1]
2460
content, _ = knit._parse_record(revision_id, raw_data)
2461
if record.storage_kind == 'knit-delta-gz':
2462
line_iterator = factory.get_linedelta_content(content)
2463
elif record.storage_kind == 'knit-ft-gz':
2464
line_iterator = factory.get_fulltext_content(content)
2465
content_text_keys.update(find_text_keys(
2466
[(line, revision_id) for line in line_iterator]))
2467
revision_keys = [(r,) for r in revision_ids]
2468
def _filtered_inv_stream():
2469
source_vf = from_repo.inventories
2470
stream = source_vf.get_record_stream(revision_keys,
2472
for record in stream:
2473
if record.storage_kind == 'absent':
2474
raise errors.NoSuchRevision(from_repo, record.key)
2475
find_text_keys_from_content(record)
2477
self._text_keys = content_text_keys - parent_text_keys
2478
return ('inventories', _filtered_inv_stream())
2480
def _get_text_stream(self):
2481
# Note: We know we don't have to handle adding root keys, because both
2482
# the source and target are the identical network name.
2483
text_stream = self.from_repository.texts.get_record_stream(
2484
self._text_keys, self._text_fetch_order, False)
2485
return ('texts', text_stream)
2487
def get_stream(self, search):
2488
revision_ids = search.get_keys()
2489
for stream_info in self._fetch_revision_texts(revision_ids):
2491
self._revision_keys = [(rev_id,) for rev_id in revision_ids]
2492
yield self._get_filtered_inv_stream(revision_ids)
2493
yield self._get_text_stream()
2497
2242
class RepositoryFormatPack(MetaDirRepositoryFormat):
2498
2243
"""Format logic for pack structured repositories.
2546
2289
utf8_files = [('format', self.get_format_string())]
2548
2291
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2549
repository = self.open(a_bzrdir=a_bzrdir, _found=True)
2550
self._run_post_repo_init_hooks(repository, a_bzrdir, shared)
2292
return self.open(a_bzrdir=a_bzrdir, _found=True)
2553
2294
def open(self, a_bzrdir, _found=False, _override_transport=None):
2554
2295
"""See RepositoryFormat.open().
2636
2379
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2381
def check_conversion_target(self, target_format):
2382
if not target_format.rich_root_data:
2383
raise errors.BadConversionTarget(
2384
'Does not support rich root data.', target_format)
2385
if not getattr(target_format, 'supports_tree_reference', False):
2386
raise errors.BadConversionTarget(
2387
'Does not support nested trees', target_format)
2638
2389
def get_format_string(self):
2639
2390
"""See RepositoryFormat.get_format_string()."""
2640
2391
return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2877
2654
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2657
class RepositoryFormatPackDevelopment2(RepositoryFormatPack):
2658
"""A no-subtrees development repository.
2660
This format should be retained until the second release after bzr 1.7.
2662
This is pack-1.6.1 with B+Tree indices.
2665
repository_class = KnitPackRepository
2666
_commit_builder_class = PackCommitBuilder
2667
supports_external_lookups = True
2668
# What index classes to use
2669
index_builder_class = BTreeBuilder
2670
index_class = BTreeGraphIndex
2671
# Set to true to get the fast-commit code path tested until a really fast
2672
# format lands in trunk. Not actually fast in this format.
2676
def _serializer(self):
2677
return xml5.serializer_v5
2679
def _get_matching_bzrdir(self):
2680
return bzrdir.format_registry.make_bzrdir('development2')
2682
def _ignore_setting_bzrdir(self, format):
2685
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2687
def get_format_string(self):
2688
"""See RepositoryFormat.get_format_string()."""
2689
return "Bazaar development format 2 (needs bzr.dev from before 1.8)\n"
2691
def get_format_description(self):
2692
"""See RepositoryFormat.get_format_description()."""
2693
return ("Development repository format, currently the same as "
2694
"1.6.1 with B+Trees.\n")
2696
def check_conversion_target(self, target_format):
2880
2700
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2881
2701
"""A subtrees development repository.
2883
2703
This format should be retained until the second release after bzr 1.7.
2885
2705
1.6.1-subtree[as it might have been] with B+Tree indices.
2887
This is [now] retained until we have a CHK based subtree format in
2891
2708
repository_class = KnitPackRepository
2892
2709
_commit_builder_class = PackRootCommitBuilder
2893
2710
rich_root_data = True
2895
2711
supports_tree_reference = True
2896
2712
supports_external_lookups = True
2897
2713
# What index classes to use
2905
2721
def _get_matching_bzrdir(self):
2906
2722
return bzrdir.format_registry.make_bzrdir(
2907
'development-subtree')
2723
'development2-subtree')
2909
2725
def _ignore_setting_bzrdir(self, format):
2912
2728
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2730
def check_conversion_target(self, target_format):
2731
if not target_format.rich_root_data:
2732
raise errors.BadConversionTarget(
2733
'Does not support rich root data.', target_format)
2734
if not getattr(target_format, 'supports_tree_reference', False):
2735
raise errors.BadConversionTarget(
2736
'Does not support nested trees', target_format)
2914
2738
def get_format_string(self):
2915
2739
"""See RepositoryFormat.get_format_string()."""
2916
2740
return ("Bazaar development format 2 with subtree support "