382
382
suitable for production use. :XXX
385
def __init__(self, transport, name, size, unlimited_cache=False):
385
def __init__(self, transport, name, size, unlimited_cache=False, offset=0):
386
386
"""Open an index called name on transport.
388
388
:param transport: A bzrlib.transport.Transport.
444
447
mutter('Reading entire index %s', self._transport.abspath(self._name))
445
448
if stream is None:
446
449
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:])
447
454
self._read_prefix(stream)
448
455
self._expected_elements = 3 + self._key_length
1190
1198
self._buffer_all()
1201
base_offset = self._base_offset
1202
if base_offset != 0:
1203
# Rewrite the ranges for the offset
1204
readv_ranges = [(start+base_offset, size)
1205
for start, size in readv_ranges]
1193
1206
readv_data = self._transport.readv(self._name, readv_ranges, True,
1207
self._size + self._base_offset)
1196
1209
for offset, data in readv_data:
1210
offset -= base_offset
1197
1211
self._bytes_read += len(data)
1213
# transport.readv() expanded to extra data which isn't part of
1215
data = data[-offset:]
1198
1217
if offset == 0 and len(data) == self._size:
1199
1218
# We read the whole range, most likely because the
1200
1219
# Transport upcast our readv ranges into one long request
1229
1248
Queries against the combined index will be made against the first index,
1230
and then the second and so on. The order of index's can thus influence
1249
and then the second and so on. The order of indices can thus influence
1231
1250
performance significantly. For example, if one index is on local disk and a
1232
1251
second on a remote server, the local disk index should be before the other
1233
1252
in the index list.
1254
Also, queries tend to need results from the same indices as previous
1255
queries. So the indices will be reordered after every query to put the
1256
indices that had the result(s) of that query first (while otherwise
1257
preserving the relative ordering).
1236
1260
def __init__(self, indices, reload_func=None):
1244
1268
self._indices = indices
1245
1269
self._reload_func = reload_func
1270
# Sibling indices are other CombinedGraphIndex that we should call
1271
# _move_to_front_by_name on when we auto-reorder ourself.
1272
self._sibling_indices = []
1273
# A list of names that corresponds to the instances in self._indices,
1274
# so _index_names[0] is always the name for _indices[0], etc. Sibling
1275
# indices must all use the same set of names as each other.
1276
self._index_names = [None] * len(self._indices)
1247
1278
def __repr__(self):
1248
1279
return "%s(%s)" % (
1272
1303
has_key = _has_key_from_parent_map
1274
def insert_index(self, pos, index):
1305
def insert_index(self, pos, index, name=None):
1275
1306
"""Insert a new index in the list of indices to query.
1277
1308
:param pos: The position to insert the index.
1278
1309
:param index: The index to insert.
1310
:param name: a name for this index, e.g. a pack name. These names can
1311
be used to reflect index reorderings to related CombinedGraphIndex
1312
instances that use the same names. (see set_sibling_indices)
1280
1314
self._indices.insert(pos, index)
1315
self._index_names.insert(pos, name)
1282
1317
def iter_all_entries(self):
1283
1318
"""Iterate over all keys within the index
1308
1343
value and are only reported once.
1310
1345
:param keys: An iterable providing the keys to be retrieved.
1311
:return: An iterable of (index, key, reference_lists, value). There is no
1312
defined order for the result iteration - it will be in the most
1346
:return: An iterable of (index, key, reference_lists, value). There is
1347
no defined order for the result iteration - it will be in the most
1313
1348
efficient order for the index.
1315
1350
keys = set(keys)
1318
1354
for index in self._indices:
1321
1358
for node in index.iter_entries(keys):
1322
1359
keys.remove(node[1])
1363
hit_indices.append(index)
1325
1365
except errors.NoSuchFile:
1326
1366
self._reload_or_raise()
1367
self._move_to_front(hit_indices)
1328
1369
def iter_entries_prefix(self, keys):
1329
1370
"""Iterate over keys within the index using prefix matching.
1351
1392
seen_keys = set()
1354
1396
for index in self._indices:
1355
1398
for node in index.iter_entries_prefix(keys):
1356
1399
if node[1] in seen_keys:
1358
1401
seen_keys.add(node[1])
1405
hit_indices.append(index)
1361
1407
except errors.NoSuchFile:
1362
1408
self._reload_or_raise()
1409
self._move_to_front(hit_indices)
1411
def _move_to_front(self, hit_indices):
1412
"""Rearrange self._indices so that hit_indices are first.
1414
Order is maintained as much as possible, e.g. the first unhit index
1415
will be the first index in _indices after the hit_indices, and the
1416
hit_indices will be present in exactly the order they are passed to
1419
_move_to_front propagates to all objects in self._sibling_indices by
1420
calling _move_to_front_by_name.
1422
if self._indices[:len(hit_indices)] == hit_indices:
1423
# The 'hit_indices' are already at the front (and in the same
1424
# order), no need to re-order
1426
hit_names = self._move_to_front_by_index(hit_indices)
1427
for sibling_idx in self._sibling_indices:
1428
sibling_idx._move_to_front_by_name(hit_names)
1430
def _move_to_front_by_index(self, hit_indices):
1431
"""Core logic for _move_to_front.
1433
Returns a list of names corresponding to the hit_indices param.
1435
indices_info = zip(self._index_names, self._indices)
1436
if 'index' in debug.debug_flags:
1437
mutter('CombinedGraphIndex reordering: currently %r, promoting %r',
1438
indices_info, hit_indices)
1441
new_hit_indices = []
1444
for offset, (name, idx) in enumerate(indices_info):
1445
if idx in hit_indices:
1446
hit_names.append(name)
1447
new_hit_indices.append(idx)
1448
if len(new_hit_indices) == len(hit_indices):
1449
# We've found all of the hit entries, everything else is
1451
unhit_names.extend(self._index_names[offset+1:])
1452
unhit_indices.extend(self._indices[offset+1:])
1455
unhit_names.append(name)
1456
unhit_indices.append(idx)
1458
self._indices = new_hit_indices + unhit_indices
1459
self._index_names = hit_names + unhit_names
1460
if 'index' in debug.debug_flags:
1461
mutter('CombinedGraphIndex reordered: %r', self._indices)
1464
def _move_to_front_by_name(self, hit_names):
1465
"""Moves indices named by 'hit_names' to front of the search order, as
1466
described in _move_to_front.
1468
# Translate names to index instances, and then call
1469
# _move_to_front_by_index.
1470
indices_info = zip(self._index_names, self._indices)
1472
for name, idx in indices_info:
1473
if name in hit_names:
1474
hit_indices.append(idx)
1475
self._move_to_front_by_index(hit_indices)
1364
1477
def find_ancestry(self, keys, ref_list_num):
1365
1478
"""Find the complete ancestry for the given set of keys.