205
205
if reference not in self._nodes:
206
206
self._check_key(reference)
207
207
absent_references.append(reference)
208
reference_list = as_st([as_st(ref).intern()
209
for ref in reference_list])
210
node_refs.append(reference_list)
208
node_refs.append(as_st(reference_list))
211
209
return as_st(node_refs), absent_references
213
211
def add_node(self, key, value, references=()):
382
380
suitable for production use. :XXX
385
def __init__(self, transport, name, size, unlimited_cache=False, offset=0):
383
def __init__(self, transport, name, size, unlimited_cache=False):
386
384
"""Open an index called name on transport.
388
386
:param transport: A bzrlib.transport.Transport.
447
442
mutter('Reading entire index %s', self._transport.abspath(self._name))
448
443
if stream is None:
449
444
stream = self._transport.get(self._name)
450
if self._base_offset != 0:
451
# This is wasteful, but it is better than dealing with
452
# adjusting all the offsets, etc.
453
stream = StringIO(stream.read()[self._base_offset:])
454
445
self._read_prefix(stream)
455
446
self._expected_elements = 3 + self._key_length
1197
1187
self._buffer_all()
1200
base_offset = self._base_offset
1201
if base_offset != 0:
1202
# Rewrite the ranges for the offset
1203
readv_ranges = [(start+base_offset, size)
1204
for start, size in readv_ranges]
1205
1190
readv_data = self._transport.readv(self._name, readv_ranges, True,
1206
self._size + self._base_offset)
1208
1193
for offset, data in readv_data:
1209
offset -= base_offset
1210
1194
self._bytes_read += len(data)
1212
# transport.readv() expanded to extra data which isn't part of
1214
data = data[-offset:]
1216
1195
if offset == 0 and len(data) == self._size:
1217
1196
# We read the whole range, most likely because the
1218
1197
# Transport upcast our readv ranges into one long request
1247
1226
Queries against the combined index will be made against the first index,
1248
and then the second and so on. The order of indices can thus influence
1227
and then the second and so on. The order of index's can thus influence
1249
1228
performance significantly. For example, if one index is on local disk and a
1250
1229
second on a remote server, the local disk index should be before the other
1251
1230
in the index list.
1253
Also, queries tend to need results from the same indices as previous
1254
queries. So the indices will be reordered after every query to put the
1255
indices that had the result(s) of that query first (while otherwise
1256
preserving the relative ordering).
1259
1233
def __init__(self, indices, reload_func=None):
1267
1241
self._indices = indices
1268
1242
self._reload_func = reload_func
1269
# Sibling indices are other CombinedGraphIndex that we should call
1270
# _move_to_front_by_name on when we auto-reorder ourself.
1271
self._sibling_indices = []
1272
# A list of names that corresponds to the instances in self._indices,
1273
# so _index_names[0] is always the name for _indices[0], etc. Sibling
1274
# indices must all use the same set of names as each other.
1275
self._index_names = [None] * len(self._indices)
1277
1244
def __repr__(self):
1278
1245
return "%s(%s)" % (
1302
1269
has_key = _has_key_from_parent_map
1304
def insert_index(self, pos, index, name=None):
1271
def insert_index(self, pos, index):
1305
1272
"""Insert a new index in the list of indices to query.
1307
1274
:param pos: The position to insert the index.
1308
1275
:param index: The index to insert.
1309
:param name: a name for this index, e.g. a pack name. These names can
1310
be used to reflect index reorderings to related CombinedGraphIndex
1311
instances that use the same names. (see set_sibling_indices)
1313
1277
self._indices.insert(pos, index)
1314
self._index_names.insert(pos, name)
1316
1279
def iter_all_entries(self):
1317
1280
"""Iterate over all keys within the index
1342
1305
value and are only reported once.
1344
1307
:param keys: An iterable providing the keys to be retrieved.
1345
:return: An iterable of (index, key, reference_lists, value). There is
1346
no defined order for the result iteration - it will be in the most
1308
:return: An iterable of (index, key, reference_lists, value). There is no
1309
defined order for the result iteration - it will be in the most
1347
1310
efficient order for the index.
1349
1312
keys = set(keys)
1353
1315
for index in self._indices:
1357
1318
for node in index.iter_entries(keys):
1358
1319
keys.remove(node[1])
1362
hit_indices.append(index)
1364
1322
except errors.NoSuchFile:
1365
1323
self._reload_or_raise()
1366
self._move_to_front(hit_indices)
1368
1325
def iter_entries_prefix(self, keys):
1369
1326
"""Iterate over keys within the index using prefix matching.
1391
1348
seen_keys = set()
1395
1351
for index in self._indices:
1397
1352
for node in index.iter_entries_prefix(keys):
1398
1353
if node[1] in seen_keys:
1400
1355
seen_keys.add(node[1])
1404
hit_indices.append(index)
1406
1358
except errors.NoSuchFile:
1407
1359
self._reload_or_raise()
1408
self._move_to_front(hit_indices)
1410
def _move_to_front(self, hit_indices):
1411
"""Rearrange self._indices so that hit_indices are first.
1413
Order is maintained as much as possible, e.g. the first unhit index
1414
will be the first index in _indices after the hit_indices, and the
1415
hit_indices will be present in exactly the order they are passed to
1418
_move_to_front propagates to all objects in self._sibling_indices by
1419
calling _move_to_front_by_name.
1421
if self._indices[:len(hit_indices)] == hit_indices:
1422
# The 'hit_indices' are already at the front (and in the same
1423
# order), no need to re-order
1425
hit_names = self._move_to_front_by_index(hit_indices)
1426
for sibling_idx in self._sibling_indices:
1427
sibling_idx._move_to_front_by_name(hit_names)
1429
def _move_to_front_by_index(self, hit_indices):
1430
"""Core logic for _move_to_front.
1432
Returns a list of names corresponding to the hit_indices param.
1434
indices_info = zip(self._index_names, self._indices)
1435
if 'index' in debug.debug_flags:
1436
mutter('CombinedGraphIndex reordering: currently %r, promoting %r',
1437
indices_info, hit_indices)
1440
new_hit_indices = []
1443
for offset, (name, idx) in enumerate(indices_info):
1444
if idx in hit_indices:
1445
hit_names.append(name)
1446
new_hit_indices.append(idx)
1447
if len(new_hit_indices) == len(hit_indices):
1448
# We've found all of the hit entries, everything else is
1450
unhit_names.extend(self._index_names[offset+1:])
1451
unhit_indices.extend(self._indices[offset+1:])
1454
unhit_names.append(name)
1455
unhit_indices.append(idx)
1457
self._indices = new_hit_indices + unhit_indices
1458
self._index_names = hit_names + unhit_names
1459
if 'index' in debug.debug_flags:
1460
mutter('CombinedGraphIndex reordered: %r', self._indices)
1463
def _move_to_front_by_name(self, hit_names):
1464
"""Moves indices named by 'hit_names' to front of the search order, as
1465
described in _move_to_front.
1467
# Translate names to index instances, and then call
1468
# _move_to_front_by_index.
1469
indices_info = zip(self._index_names, self._indices)
1471
for name, idx in indices_info:
1472
if name in hit_names:
1473
hit_indices.append(idx)
1474
self._move_to_front_by_index(hit_indices)
1476
1361
def find_ancestry(self, keys, ref_list_num):
1477
1362
"""Find the complete ancestry for the given set of keys.