227
228
return self.index_name('text', name)
229
230
def _replace_index_with_readonly(self, index_type):
230
unlimited_cache = False
231
if index_type == 'chk':
232
unlimited_cache = True
233
231
setattr(self, index_type + '_index',
234
232
self.index_class(self.index_transport,
235
233
self.index_name(index_type, self.name),
236
self.index_sizes[self.index_offset(index_type)],
237
unlimited_cache=unlimited_cache))
234
self.index_sizes[self.index_offset(index_type)]))
240
237
class ExistingPack(Pack):
588
578
flush_func=flush_func)
589
579
self.add_callback = None
581
def replace_indices(self, index_to_pack, indices):
582
"""Replace the current mappings with fresh ones.
584
This should probably not be used eventually, rather incremental add and
585
removal of indices. It has been added during refactoring of existing
588
:param index_to_pack: A mapping from index objects to
589
(transport, name) tuples for the pack file data.
590
:param indices: A list of indices.
592
# refresh the revision pack map dict without replacing the instance.
593
self.index_to_pack.clear()
594
self.index_to_pack.update(index_to_pack)
595
# XXX: API break - clearly a 'replace' method would be good?
596
self.combined_index._indices[:] = indices
597
# the current add nodes callback for the current writable index if
599
self.add_callback = None
591
601
def add_index(self, index, pack):
592
602
"""Add index to the aggregate, which is an index for Pack pack.
626
636
self.data_access.set_writer(None, None, (None, None))
627
637
self.index_to_pack.clear()
628
638
del self.combined_index._indices[:]
629
del self.combined_index._index_names[:]
630
639
self.add_callback = None
632
def remove_index(self, index):
641
def remove_index(self, index, pack):
633
642
"""Remove index from the indices used to answer queries.
635
644
:param index: An index from the pack parameter.
645
:param pack: A Pack instance.
637
647
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]
648
self.combined_index._indices.remove(index)
641
649
if (self.add_callback is not None and
642
650
getattr(index, 'add_nodes', None) == self.add_callback):
643
651
self.add_callback = None
1399
1407
self.inventory_index = AggregateIndex(self.reload_pack_names, flush)
1400
1408
self.text_index = AggregateIndex(self.reload_pack_names, flush)
1401
1409
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]
1404
1410
if use_chk_index:
1405
1411
self.chk_index = AggregateIndex(self.reload_pack_names, flush)
1406
all_indices.append(self.chk_index)
1408
1413
# used to determine if we're using a chk_index elsewhere.
1409
1414
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]))
1416
1415
# resumed packs
1417
1416
self._resumed_packs = []
1420
return '%s(%r)' % (self.__class__.__name__, self.repo)
1422
1418
def add_pack_to_memory(self, pack):
1423
1419
"""Make a Pack object available to the repository to satisfy queries.
1505
1501
'containing %d revisions. Packing %d files into %d affecting %d'
1506
1502
' revisions', self, total_packs, total_revisions, num_old_packs,
1507
1503
num_new_packs, num_revs_affected)
1508
result = self._execute_pack_operations(pack_operations,
1504
self._execute_pack_operations(pack_operations,
1509
1505
reload_func=self._restart_autopack)
1510
1506
mutter('Auto-packing repository %s completed', self)
1513
1509
def _execute_pack_operations(self, pack_operations, _packer_class=Packer,
1514
1510
reload_func=None):
1538
1534
self._remove_pack_from_memory(pack)
1539
1535
# 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)
1537
self._save_pack_names(clear_obsolete_packs=True)
1538
# Move the old packs out of the way now they are no longer referenced.
1539
for revision_count, packs in pack_operations:
1540
self._obsolete_packs(packs)
1548
1542
def _flush_new_pack(self):
1549
1543
if self._new_pack is not None:
1560
1554
def _already_packed(self):
1561
1555
"""Is the collection already packed?"""
1562
return not (self.repo._format.pack_compresses or (len(self._names) > 1))
1556
return len(self._names) < 2
1564
def pack(self, hint=None, clean_obsolete_packs=False):
1565
1559
"""Pack the pack collection totally."""
1566
1560
self.ensure_loaded()
1567
1561
total_packs = len(self._names)
1568
1562
if self._already_packed():
1563
# This is arguably wrong because we might not be optimal, but for
1564
# now lets leave it in. (e.g. reconcile -> one pack. But not
1570
1567
total_revisions = self.revision_index.combined_index.key_count()
1571
1568
# XXX: the following may want to be a class, to pack with a given
1573
1570
mutter('Packing repository %s, which has %d pack files, '
1574
'containing %d revisions with hint %r.', self, total_packs,
1575
total_revisions, hint)
1571
'containing %d revisions into 1 packs.', self, total_packs,
1576
1573
# determine which packs need changing
1574
pack_distribution = [1]
1577
1575
pack_operations = [[0, []]]
1578
1576
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)
1577
pack_operations[-1][0] += pack.get_revision_count()
1578
pack_operations[-1][1].append(pack)
1584
1579
self._execute_pack_operations(pack_operations, OptimisingPacker)
1586
if clean_obsolete_packs:
1587
self._clear_obsolete_packs()
1589
1581
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1590
1582
"""Plan a pack operation.
1704
1696
txt_index = self._make_index(name, '.tix', resume=True)
1705
1697
sig_index = self._make_index(name, '.six', resume=True)
1706
1698
if self.chk_index is not None:
1707
chk_index = self._make_index(name, '.cix', resume=True,
1708
unlimited_cache=True)
1699
chk_index = self._make_index(name, '.cix', resume=True)
1710
1701
chk_index = None
1711
1702
result = self.resumed_pack_factory(name, rev_index, inv_index,
1785
1775
:param return: None.
1787
1777
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"
1778
pack.pack_transport.rename(pack.file_name(),
1779
'../obsolete_packs/' + pack.file_name())
1795
1780
# TODO: Probably needs to know all possible indices for this pack
1796
1781
# - or maybe list the directory and move all indices matching this
1797
1782
# name whether we recognize it or not?
1799
1784
if self.chk_index is not None:
1800
1785
suffixes.append('.cix')
1801
1786
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"
1787
self._index_transport.rename(pack.name + suffix,
1788
'../obsolete_packs/' + pack.name + suffix)
1809
1790
def pack_distribution(self, total_revisions):
1810
1791
"""Generate a list of the number of revisions to put in each pack.
1836
1817
self._remove_pack_indices(pack)
1837
1818
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)
1820
def _remove_pack_indices(self, pack):
1821
"""Remove the indices for pack from the aggregated indices."""
1822
self.revision_index.remove_index(pack.revision_index, pack)
1823
self.inventory_index.remove_index(pack.inventory_index, pack)
1824
self.text_index.remove_index(pack.text_index, pack)
1825
self.signature_index.remove_index(pack.signature_index, pack)
1826
if self.chk_index is not None:
1827
self.chk_index.remove_index(pack.chk_index, pack)
1856
1829
def reset(self):
1857
1830
"""Clear all cached data."""
1966
1938
:param clear_obsolete_packs: If True, clear out the contents of the
1967
1939
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 = []
1973
1941
self.lock_names()
1975
1943
builder = self._index_builder_class()
1976
(disk_nodes, deleted_nodes, new_nodes,
1977
orig_disk_nodes) = self._diff_pack_names()
1944
disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1978
1945
# TODO: handle same-name, index-size-changes here -
1979
1946
# e.g. use the value from disk, not ours, *unless* we're the one
1982
1949
builder.add_node(key, value)
1983
1950
self.transport.put_file('pack-names', builder.finish(),
1984
1951
mode=self.repo.bzrdir._get_file_mode())
1952
# move the baseline forward
1985
1953
self._packs_at_load = disk_nodes
1986
1954
if clear_obsolete_packs:
1989
to_preserve = set([o.name for o in obsolete_packs])
1990
already_obsolete = self._clear_obsolete_packs(to_preserve)
1955
self._clear_obsolete_packs()
1992
1957
self._unlock_names()
1993
1958
# synchronise the memory packs list with what we just wrote:
1994
1959
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
1961
def reload_pack_names(self):
2007
1962
"""Sync our pack listing with what is present in the repository.
2023
1978
# 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
1979
disk_nodes, _, _ = self._diff_pack_names()
1980
self._packs_at_load = disk_nodes
2030
1981
(removed, added,
2031
1982
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
2032
1983
if removed or added or modified:
2042
1993
raise errors.RetryAutopack(self.repo, False, sys.exc_info())
2044
def _clear_obsolete_packs(self, preserve=None):
1995
def _clear_obsolete_packs(self):
2045
1996
"""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
1998
obsolete_pack_transport = self.transport.clone('obsolete_packs')
2052
if preserve is None:
2054
1999
for filename in obsolete_pack_transport.list_dir('.'):
2055
name, ext = osutils.splitext(filename)
2058
if name in preserve:
2061
2001
obsolete_pack_transport.delete(filename)
2062
2002
except (errors.PathError, errors.TransportError), e:
2063
warning("couldn't delete obsolete pack, skipping it:\n%s"
2003
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
2067
2005
def _start_write_group(self):
2068
2006
# Do not permit preparation for writing if we're not in a 'write lock'.
2095
2033
# FIXME: just drop the transient index.
2096
2034
# forget what names there are
2097
2035
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()
2037
self._new_pack.abort()
2039
# XXX: If we aborted while in the middle of finishing the write
2040
# group, _remove_pack_indices can fail because the indexes are
2041
# already gone. If they're not there we shouldn't fail in this
2042
# case. -- mbp 20081113
2043
self._remove_pack_indices(self._new_pack)
2044
self._new_pack = None
2107
2045
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()
2047
resumed_pack.abort()
2049
# See comment in previous finally block.
2051
self._remove_pack_indices(resumed_pack)
2113
2054
del self._resumed_packs[:]
2115
2056
def _remove_resumed_pack_indices(self):
2141
2072
raise errors.BzrCheckError(
2142
2073
"Repository %s has missing compression parent(s) %r "
2143
2074
% (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
2075
self._remove_pack_indices(self._new_pack)
2150
any_new_content = False
2076
should_autopack = False
2151
2077
if self._new_pack.data_inserted():
2152
2078
# get all the data to disk and read to use
2153
2079
self._new_pack.finish()
2154
2080
self.allocate(self._new_pack)
2155
2081
self._new_pack = None
2156
any_new_content = True
2082
should_autopack = True
2158
2084
self._new_pack.abort()
2159
2085
self._new_pack = None
2281
2204
self._reconcile_fixes_text_parents = True
2282
2205
self._reconcile_backsup_inventory = False
2284
def _warn_if_deprecated(self, branch=None):
2207
def _warn_if_deprecated(self):
2285
2208
# This class isn't deprecated, but one sub-format is
2286
2209
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
2287
super(KnitPackRepository, self)._warn_if_deprecated(branch)
2210
from bzrlib import repository
2211
if repository._deprecation_warning_done:
2213
repository._deprecation_warning_done = True
2214
warning("Format %s for %s is deprecated - please use"
2215
" 'bzr upgrade --1.6.1-rich-root'"
2216
% (self._format, self.bzrdir.transport.base))
2289
2218
def _abort_write_group(self):
2290
self.revisions._index._key_dependencies.clear()
2219
self.revisions._index._key_dependencies.refs.clear()
2291
2220
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)
2222
def _find_inconsistent_revision_parents(self):
2223
"""Find revisions with incorrectly cached parents.
2225
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
2226
parents-in-revision).
2228
if not self.is_locked():
2229
raise errors.ObjectNotLocked(self)
2230
pb = ui.ui_factory.nested_progress_bar()
2233
revision_nodes = self._pack_collection.revision_index \
2234
.combined_index.iter_all_entries()
2235
index_positions = []
2236
# Get the cached index values for all revisions, and also the
2237
# location in each index of the revision text so we can perform
2239
for index, key, value, refs in revision_nodes:
2240
node = (index, key, value, refs)
2241
index_memo = self.revisions._index._node_to_position(node)
2242
if index_memo[0] != index:
2243
raise AssertionError('%r != %r' % (index_memo[0], index))
2244
index_positions.append((index_memo, key[0],
2245
tuple(parent[0] for parent in refs[0])))
2246
pb.update("Reading revision index", 0, 0)
2247
index_positions.sort()
2249
pb.update("Checking cached revision graph", 0,
2250
len(index_positions))
2251
for offset in xrange(0, len(index_positions), 1000):
2252
pb.update("Checking cached revision graph", offset)
2253
to_query = index_positions[offset:offset + batch_size]
2256
rev_ids = [item[1] for item in to_query]
2257
revs = self.get_revisions(rev_ids)
2258
for revision, item in zip(revs, to_query):
2259
index_parents = item[2]
2260
rev_parents = tuple(revision.parent_ids)
2261
if index_parents != rev_parents:
2262
result.append((revision.revision_id, index_parents,
2298
2268
def _make_parents_provider(self):
2299
2269
return graph.CachingParentsProvider(self)
2307
2277
self._pack_collection._start_write_group()
2309
2279
def _commit_write_group(self):
2310
hint = self._pack_collection._commit_write_group()
2311
self.revisions._index._key_dependencies.clear()
2280
self.revisions._index._key_dependencies.refs.clear()
2281
return self._pack_collection._commit_write_group()
2314
2283
def suspend_write_group(self):
2315
2284
# XXX check self._write_group is self.get_transaction()?
2316
2285
tokens = self._pack_collection._suspend_write_group()
2317
self.revisions._index._key_dependencies.clear()
2286
self.revisions._index._key_dependencies.refs.clear()
2318
2287
self._write_group = None
2321
2290
def _resume_write_group(self, tokens):
2322
2291
self._start_write_group()
2324
self._pack_collection._resume_write_group(tokens)
2325
except errors.UnresumableWriteGroup:
2326
self._abort_write_group()
2292
self._pack_collection._resume_write_group(tokens)
2328
2293
for pack in self._pack_collection._resumed_packs:
2329
2294
self.revisions._index.scan_unvalidated_index(pack.revision_index)
2381
2338
raise NotImplementedError(self.dont_leave_lock_in_place)
2383
2340
@needs_write_lock
2384
def pack(self, hint=None, clean_obsolete_packs=False):
2385
2342
"""Compress the data within the repository.
2387
2344
This will pack all the data to a single pack. In future it may
2388
2345
recompress deltas or do other such expensive operations.
2390
self._pack_collection.pack(hint=hint, clean_obsolete_packs=clean_obsolete_packs)
2347
self._pack_collection.pack()
2392
2349
@needs_write_lock
2393
2350
def reconcile(self, other=None, thorough=False):
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
2383
class RepositoryFormatPack(MetaDirRepositoryFormat):
2501
2384
"""Format logic for pack structured repositories.
2549
2432
utf8_files = [('format', self.get_format_string())]
2551
2434
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2552
repository = self.open(a_bzrdir=a_bzrdir, _found=True)
2553
self._run_post_repo_init_hooks(repository, a_bzrdir, shared)
2435
return self.open(a_bzrdir=a_bzrdir, _found=True)
2556
2437
def open(self, a_bzrdir, _found=False, _override_transport=None):
2557
2438
"""See RepositoryFormat.open().
2639
2522
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2524
def check_conversion_target(self, target_format):
2525
if not target_format.rich_root_data:
2526
raise errors.BadConversionTarget(
2527
'Does not support rich root data.', target_format)
2528
if not getattr(target_format, 'supports_tree_reference', False):
2529
raise errors.BadConversionTarget(
2530
'Does not support nested trees', target_format)
2641
2532
def get_format_string(self):
2642
2533
"""See RepositoryFormat.get_format_string()."""
2643
2534
return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2915
2831
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2833
def check_conversion_target(self, target_format):
2834
if not target_format.rich_root_data:
2835
raise errors.BadConversionTarget(
2836
'Does not support rich root data.', target_format)
2837
if not getattr(target_format, 'supports_tree_reference', False):
2838
raise errors.BadConversionTarget(
2839
'Does not support nested trees', target_format)
2917
2841
def get_format_string(self):
2918
2842
"""See RepositoryFormat.get_format_string()."""
2919
2843
return ("Bazaar development format 2 with subtree support "