/usr/lib/python2.7/dist-packages/bzrlib/chk_map.py is in python-bzrlib 2.7.0-2ubuntu3.1.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 | # Copyright (C) 2008-2011 Canonical Ltd
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
"""Persistent maps from tuple_of_strings->string using CHK stores.
Overview and current status:
The CHKMap class implements a dict from tuple_of_strings->string by using a trie
with internal nodes of 8-bit fan out; The key tuples are mapped to strings by
joining them by \x00, and \x00 padding shorter keys out to the length of the
longest key. Leaf nodes are packed as densely as possible, and internal nodes
are all an additional 8-bits wide leading to a sparse upper tree.
Updates to a CHKMap are done preferentially via the apply_delta method, to
allow optimisation of the update operation; but individual map/unmap calls are
possible and supported. Individual changes via map/unmap are buffered in memory
until the _save method is called to force serialisation of the tree.
apply_delta records its changes immediately by performing an implicit _save.
TODO:
-----
Densely packed upper nodes.
"""
from __future__ import absolute_import
import heapq
import threading
from bzrlib import lazy_import
lazy_import.lazy_import(globals(), """
from bzrlib import (
errors,
)
""")
from bzrlib import (
errors,
lru_cache,
osutils,
registry,
static_tuple,
trace,
)
from bzrlib.static_tuple import StaticTuple
# approx 4MB
# If each line is 50 bytes, and you have 255 internal pages, with 255-way fan
# out, it takes 3.1MB to cache the layer.
_PAGE_CACHE_SIZE = 4*1024*1024
# Per thread caches for 2 reasons:
# - in the server we may be serving very different content, so we get less
# cache thrashing.
# - we avoid locking on every cache lookup.
_thread_caches = threading.local()
# The page cache.
_thread_caches.page_cache = None
def _get_cache():
"""Get the per-thread page cache.
We need a function to do this because in a new thread the _thread_caches
threading.local object does not have the cache initialized yet.
"""
page_cache = getattr(_thread_caches, 'page_cache', None)
if page_cache is None:
# We are caching bytes so len(value) is perfectly accurate
page_cache = lru_cache.LRUSizeCache(_PAGE_CACHE_SIZE)
_thread_caches.page_cache = page_cache
return page_cache
def clear_cache():
_get_cache().clear()
# If a ChildNode falls below this many bytes, we check for a remap
_INTERESTING_NEW_SIZE = 50
# If a ChildNode shrinks by more than this amount, we check for a remap
_INTERESTING_SHRINKAGE_LIMIT = 20
def _search_key_plain(key):
"""Map the key tuple into a search string that just uses the key bytes."""
return '\x00'.join(key)
search_key_registry = registry.Registry()
search_key_registry.register('plain', _search_key_plain)
class CHKMap(object):
"""A persistent map from string to string backed by a CHK store."""
__slots__ = ('_store', '_root_node', '_search_key_func')
def __init__(self, store, root_key, search_key_func=None):
"""Create a CHKMap object.
:param store: The store the CHKMap is stored in.
:param root_key: The root key of the map. None to create an empty
CHKMap.
:param search_key_func: A function mapping a key => bytes. These bytes
are then used by the internal nodes to split up leaf nodes into
multiple pages.
"""
self._store = store
if search_key_func is None:
search_key_func = _search_key_plain
self._search_key_func = search_key_func
if root_key is None:
self._root_node = LeafNode(search_key_func=search_key_func)
else:
self._root_node = self._node_key(root_key)
def apply_delta(self, delta):
"""Apply a delta to the map.
:param delta: An iterable of old_key, new_key, new_value tuples.
If new_key is not None, then new_key->new_value is inserted
into the map; if old_key is not None, then the old mapping
of old_key is removed.
"""
has_deletes = False
# Check preconditions first.
as_st = StaticTuple.from_sequence
new_items = set([as_st(key) for (old, key, value) in delta
if key is not None and old is None])
existing_new = list(self.iteritems(key_filter=new_items))
if existing_new:
raise errors.InconsistentDeltaDelta(delta,
"New items are already in the map %r." % existing_new)
# Now apply changes.
for old, new, value in delta:
if old is not None and old != new:
self.unmap(old, check_remap=False)
has_deletes = True
for old, new, value in delta:
if new is not None:
self.map(new, value)
if has_deletes:
self._check_remap()
return self._save()
def _ensure_root(self):
"""Ensure that the root node is an object not a key."""
if type(self._root_node) is StaticTuple:
# Demand-load the root
self._root_node = self._get_node(self._root_node)
def _get_node(self, node):
"""Get a node.
Note that this does not update the _items dict in objects containing a
reference to this node. As such it does not prevent subsequent IO being
performed.
:param node: A tuple key or node object.
:return: A node object.
"""
if type(node) is StaticTuple:
bytes = self._read_bytes(node)
return _deserialise(bytes, node,
search_key_func=self._search_key_func)
else:
return node
def _read_bytes(self, key):
try:
return _get_cache()[key]
except KeyError:
stream = self._store.get_record_stream([key], 'unordered', True)
bytes = stream.next().get_bytes_as('fulltext')
_get_cache()[key] = bytes
return bytes
def _dump_tree(self, include_keys=False):
"""Return the tree in a string representation."""
self._ensure_root()
res = self._dump_tree_node(self._root_node, prefix='', indent='',
include_keys=include_keys)
res.append('') # Give a trailing '\n'
return '\n'.join(res)
def _dump_tree_node(self, node, prefix, indent, include_keys=True):
"""For this node and all children, generate a string representation."""
result = []
if not include_keys:
key_str = ''
else:
node_key = node.key()
if node_key is not None:
key_str = ' %s' % (node_key[0],)
else:
key_str = ' None'
result.append('%s%r %s%s' % (indent, prefix, node.__class__.__name__,
key_str))
if type(node) is InternalNode:
# Trigger all child nodes to get loaded
list(node._iter_nodes(self._store))
for prefix, sub in sorted(node._items.iteritems()):
result.extend(self._dump_tree_node(sub, prefix, indent + ' ',
include_keys=include_keys))
else:
for key, value in sorted(node._items.iteritems()):
# Don't use prefix nor indent here to line up when used in
# tests in conjunction with assertEqualDiff
result.append(' %r %r' % (tuple(key), value))
return result
@classmethod
def from_dict(klass, store, initial_value, maximum_size=0, key_width=1,
search_key_func=None):
"""Create a CHKMap in store with initial_value as the content.
:param store: The store to record initial_value in, a VersionedFiles
object with 1-tuple keys supporting CHK key generation.
:param initial_value: A dict to store in store. Its keys and values
must be bytestrings.
:param maximum_size: The maximum_size rule to apply to nodes. This
determines the size at which no new data is added to a single node.
:param key_width: The number of elements in each key_tuple being stored
in this map.
:param search_key_func: A function mapping a key => bytes. These bytes
are then used by the internal nodes to split up leaf nodes into
multiple pages.
:return: The root chk of the resulting CHKMap.
"""
root_key = klass._create_directly(store, initial_value,
maximum_size=maximum_size, key_width=key_width,
search_key_func=search_key_func)
if type(root_key) is not StaticTuple:
raise AssertionError('we got a %s instead of a StaticTuple'
% (type(root_key),))
return root_key
@classmethod
def _create_via_map(klass, store, initial_value, maximum_size=0,
key_width=1, search_key_func=None):
result = klass(store, None, search_key_func=search_key_func)
result._root_node.set_maximum_size(maximum_size)
result._root_node._key_width = key_width
delta = []
for key, value in initial_value.items():
delta.append((None, key, value))
root_key = result.apply_delta(delta)
return root_key
@classmethod
def _create_directly(klass, store, initial_value, maximum_size=0,
key_width=1, search_key_func=None):
node = LeafNode(search_key_func=search_key_func)
node.set_maximum_size(maximum_size)
node._key_width = key_width
as_st = StaticTuple.from_sequence
node._items = dict([(as_st(key), val) for key, val
in initial_value.iteritems()])
node._raw_size = sum([node._key_value_len(key, value)
for key,value in node._items.iteritems()])
node._len = len(node._items)
node._compute_search_prefix()
node._compute_serialised_prefix()
if (node._len > 1
and maximum_size
and node._current_size() > maximum_size):
prefix, node_details = node._split(store)
if len(node_details) == 1:
raise AssertionError('Failed to split using node._split')
node = InternalNode(prefix, search_key_func=search_key_func)
node.set_maximum_size(maximum_size)
node._key_width = key_width
for split, subnode in node_details:
node.add_node(split, subnode)
keys = list(node.serialise(store))
return keys[-1]
def iter_changes(self, basis):
"""Iterate over the changes between basis and self.
:return: An iterator of tuples: (key, old_value, new_value). Old_value
is None for keys only in self; new_value is None for keys only in
basis.
"""
# Overview:
# Read both trees in lexographic, highest-first order.
# Any identical nodes we skip
# Any unique prefixes we output immediately.
# values in a leaf node are treated as single-value nodes in the tree
# which allows them to be not-special-cased. We know to output them
# because their value is a string, not a key(tuple) or node.
#
# corner cases to beware of when considering this function:
# *) common references are at different heights.
# consider two trees:
# {'a': LeafNode={'aaa':'foo', 'aab':'bar'}, 'b': LeafNode={'b'}}
# {'a': InternalNode={'aa':LeafNode={'aaa':'foo', 'aab':'bar'},
# 'ab':LeafNode={'ab':'bar'}}
# 'b': LeafNode={'b'}}
# the node with aaa/aab will only be encountered in the second tree
# after reading the 'a' subtree, but it is encountered in the first
# tree immediately. Variations on this may have read internal nodes
# like this. we want to cut the entire pending subtree when we
# realise we have a common node. For this we use a list of keys -
# the path to a node - and check the entire path is clean as we
# process each item.
if self._node_key(self._root_node) == self._node_key(basis._root_node):
return
self._ensure_root()
basis._ensure_root()
excluded_keys = set()
self_node = self._root_node
basis_node = basis._root_node
# A heap, each element is prefix, node(tuple/NodeObject/string),
# key_path (a list of tuples, tail-sharing down the tree.)
self_pending = []
basis_pending = []
def process_node(node, path, a_map, pending):
# take a node and expand it
node = a_map._get_node(node)
if type(node) == LeafNode:
path = (node._key, path)
for key, value in node._items.items():
# For a LeafNode, the key is a serialized_key, rather than
# a search_key, but the heap is using search_keys
search_key = node._search_key_func(key)
heapq.heappush(pending, (search_key, key, value, path))
else:
# type(node) == InternalNode
path = (node._key, path)
for prefix, child in node._items.items():
heapq.heappush(pending, (prefix, None, child, path))
def process_common_internal_nodes(self_node, basis_node):
self_items = set(self_node._items.items())
basis_items = set(basis_node._items.items())
path = (self_node._key, None)
for prefix, child in self_items - basis_items:
heapq.heappush(self_pending, (prefix, None, child, path))
path = (basis_node._key, None)
for prefix, child in basis_items - self_items:
heapq.heappush(basis_pending, (prefix, None, child, path))
def process_common_leaf_nodes(self_node, basis_node):
self_items = set(self_node._items.items())
basis_items = set(basis_node._items.items())
path = (self_node._key, None)
for key, value in self_items - basis_items:
prefix = self._search_key_func(key)
heapq.heappush(self_pending, (prefix, key, value, path))
path = (basis_node._key, None)
for key, value in basis_items - self_items:
prefix = basis._search_key_func(key)
heapq.heappush(basis_pending, (prefix, key, value, path))
def process_common_prefix_nodes(self_node, self_path,
basis_node, basis_path):
# Would it be more efficient if we could request both at the same
# time?
self_node = self._get_node(self_node)
basis_node = basis._get_node(basis_node)
if (type(self_node) == InternalNode
and type(basis_node) == InternalNode):
# Matching internal nodes
process_common_internal_nodes(self_node, basis_node)
elif (type(self_node) == LeafNode
and type(basis_node) == LeafNode):
process_common_leaf_nodes(self_node, basis_node)
else:
process_node(self_node, self_path, self, self_pending)
process_node(basis_node, basis_path, basis, basis_pending)
process_common_prefix_nodes(self_node, None, basis_node, None)
self_seen = set()
basis_seen = set()
excluded_keys = set()
def check_excluded(key_path):
# Note that this is N^2, it depends on us trimming trees
# aggressively to not become slow.
# A better implementation would probably have a reverse map
# back to the children of a node, and jump straight to it when
# a common node is detected, the proceed to remove the already
# pending children. bzrlib.graph has a searcher module with a
# similar problem.
while key_path is not None:
key, key_path = key_path
if key in excluded_keys:
return True
return False
loop_counter = 0
while self_pending or basis_pending:
loop_counter += 1
if not self_pending:
# self is exhausted: output remainder of basis
for prefix, key, node, path in basis_pending:
if check_excluded(path):
continue
node = basis._get_node(node)
if key is not None:
# a value
yield (key, node, None)
else:
# subtree - fastpath the entire thing.
for key, value in node.iteritems(basis._store):
yield (key, value, None)
return
elif not basis_pending:
# basis is exhausted: output remainder of self.
for prefix, key, node, path in self_pending:
if check_excluded(path):
continue
node = self._get_node(node)
if key is not None:
# a value
yield (key, None, node)
else:
# subtree - fastpath the entire thing.
for key, value in node.iteritems(self._store):
yield (key, None, value)
return
else:
# XXX: future optimisation - yield the smaller items
# immediately rather than pushing everything on/off the
# heaps. Applies to both internal nodes and leafnodes.
if self_pending[0][0] < basis_pending[0][0]:
# expand self
prefix, key, node, path = heapq.heappop(self_pending)
if check_excluded(path):
continue
if key is not None:
# a value
yield (key, None, node)
else:
process_node(node, path, self, self_pending)
continue
elif self_pending[0][0] > basis_pending[0][0]:
# expand basis
prefix, key, node, path = heapq.heappop(basis_pending)
if check_excluded(path):
continue
if key is not None:
# a value
yield (key, node, None)
else:
process_node(node, path, basis, basis_pending)
continue
else:
# common prefix: possibly expand both
if self_pending[0][1] is None:
# process next self
read_self = True
else:
read_self = False
if basis_pending[0][1] is None:
# process next basis
read_basis = True
else:
read_basis = False
if not read_self and not read_basis:
# compare a common value
self_details = heapq.heappop(self_pending)
basis_details = heapq.heappop(basis_pending)
if self_details[2] != basis_details[2]:
yield (self_details[1],
basis_details[2], self_details[2])
continue
# At least one side wasn't a simple value
if (self._node_key(self_pending[0][2]) ==
self._node_key(basis_pending[0][2])):
# Identical pointers, skip (and don't bother adding to
# excluded, it won't turn up again.
heapq.heappop(self_pending)
heapq.heappop(basis_pending)
continue
# Now we need to expand this node before we can continue
if read_self and read_basis:
# Both sides start with the same prefix, so process
# them in parallel
self_prefix, _, self_node, self_path = heapq.heappop(
self_pending)
basis_prefix, _, basis_node, basis_path = heapq.heappop(
basis_pending)
if self_prefix != basis_prefix:
raise AssertionError(
'%r != %r' % (self_prefix, basis_prefix))
process_common_prefix_nodes(
self_node, self_path,
basis_node, basis_path)
continue
if read_self:
prefix, key, node, path = heapq.heappop(self_pending)
if check_excluded(path):
continue
process_node(node, path, self, self_pending)
if read_basis:
prefix, key, node, path = heapq.heappop(basis_pending)
if check_excluded(path):
continue
process_node(node, path, basis, basis_pending)
# print loop_counter
def iteritems(self, key_filter=None):
"""Iterate over the entire CHKMap's contents."""
self._ensure_root()
if key_filter is not None:
as_st = StaticTuple.from_sequence
key_filter = [as_st(key) for key in key_filter]
return self._root_node.iteritems(self._store, key_filter=key_filter)
def key(self):
"""Return the key for this map."""
if type(self._root_node) is StaticTuple:
return self._root_node
else:
return self._root_node._key
def __len__(self):
self._ensure_root()
return len(self._root_node)
def map(self, key, value):
"""Map a key tuple to value.
:param key: A key to map.
:param value: The value to assign to key.
"""
key = StaticTuple.from_sequence(key)
# Need a root object.
self._ensure_root()
prefix, node_details = self._root_node.map(self._store, key, value)
if len(node_details) == 1:
self._root_node = node_details[0][1]
else:
self._root_node = InternalNode(prefix,
search_key_func=self._search_key_func)
self._root_node.set_maximum_size(node_details[0][1].maximum_size)
self._root_node._key_width = node_details[0][1]._key_width
for split, node in node_details:
self._root_node.add_node(split, node)
def _node_key(self, node):
"""Get the key for a node whether it's a tuple or node."""
if type(node) is tuple:
node = StaticTuple.from_sequence(node)
if type(node) is StaticTuple:
return node
else:
return node._key
def unmap(self, key, check_remap=True):
"""remove key from the map."""
key = StaticTuple.from_sequence(key)
self._ensure_root()
if type(self._root_node) is InternalNode:
unmapped = self._root_node.unmap(self._store, key,
check_remap=check_remap)
else:
unmapped = self._root_node.unmap(self._store, key)
self._root_node = unmapped
def _check_remap(self):
"""Check if nodes can be collapsed."""
self._ensure_root()
if type(self._root_node) is InternalNode:
self._root_node = self._root_node._check_remap(self._store)
def _save(self):
"""Save the map completely.
:return: The key of the root node.
"""
if type(self._root_node) is StaticTuple:
# Already saved.
return self._root_node
keys = list(self._root_node.serialise(self._store))
return keys[-1]
class Node(object):
"""Base class defining the protocol for CHK Map nodes.
:ivar _raw_size: The total size of the serialized key:value data, before
adding the header bytes, and without prefix compression.
"""
__slots__ = ('_key', '_len', '_maximum_size', '_key_width',
'_raw_size', '_items', '_search_prefix', '_search_key_func'
)
def __init__(self, key_width=1):
"""Create a node.
:param key_width: The width of keys for this node.
"""
self._key = None
# Current number of elements
self._len = 0
self._maximum_size = 0
self._key_width = key_width
# current size in bytes
self._raw_size = 0
# The pointers/values this node has - meaning defined by child classes.
self._items = {}
# The common search prefix
self._search_prefix = None
def __repr__(self):
items_str = str(sorted(self._items))
if len(items_str) > 20:
items_str = items_str[:16] + '...]'
return '%s(key:%s len:%s size:%s max:%s prefix:%s items:%s)' % (
self.__class__.__name__, self._key, self._len, self._raw_size,
self._maximum_size, self._search_prefix, items_str)
def key(self):
return self._key
def __len__(self):
return self._len
@property
def maximum_size(self):
"""What is the upper limit for adding references to a node."""
return self._maximum_size
def set_maximum_size(self, new_size):
"""Set the size threshold for nodes.
:param new_size: The size at which no data is added to a node. 0 for
unlimited.
"""
self._maximum_size = new_size
@classmethod
def common_prefix(cls, prefix, key):
"""Given 2 strings, return the longest prefix common to both.
:param prefix: This has been the common prefix for other keys, so it is
more likely to be the common prefix in this case as well.
:param key: Another string to compare to
"""
if key.startswith(prefix):
return prefix
pos = -1
# Is there a better way to do this?
for pos, (left, right) in enumerate(zip(prefix, key)):
if left != right:
pos -= 1
break
common = prefix[:pos+1]
return common
@classmethod
def common_prefix_for_keys(cls, keys):
"""Given a list of keys, find their common prefix.
:param keys: An iterable of strings.
:return: The longest common prefix of all keys.
"""
common_prefix = None
for key in keys:
if common_prefix is None:
common_prefix = key
continue
common_prefix = cls.common_prefix(common_prefix, key)
if not common_prefix:
# if common_prefix is the empty string, then we know it won't
# change further
return ''
return common_prefix
# Singleton indicating we have not computed _search_prefix yet
_unknown = object()
class LeafNode(Node):
"""A node containing actual key:value pairs.
:ivar _items: A dict of key->value items. The key is in tuple form.
:ivar _size: The number of bytes that would be used by serializing all of
the key/value pairs.
"""
__slots__ = ('_common_serialised_prefix',)
def __init__(self, search_key_func=None):
Node.__init__(self)
# All of the keys in this leaf node share this common prefix
self._common_serialised_prefix = None
if search_key_func is None:
self._search_key_func = _search_key_plain
else:
self._search_key_func = search_key_func
def __repr__(self):
items_str = str(sorted(self._items))
if len(items_str) > 20:
items_str = items_str[:16] + '...]'
return \
'%s(key:%s len:%s size:%s max:%s prefix:%s keywidth:%s items:%s)' \
% (self.__class__.__name__, self._key, self._len, self._raw_size,
self._maximum_size, self._search_prefix, self._key_width, items_str)
def _current_size(self):
"""Answer the current serialised size of this node.
This differs from self._raw_size in that it includes the bytes used for
the header.
"""
if self._common_serialised_prefix is None:
bytes_for_items = 0
prefix_len = 0
else:
# We will store a single string with the common prefix
# And then that common prefix will not be stored in any of the
# entry lines
prefix_len = len(self._common_serialised_prefix)
bytes_for_items = (self._raw_size - (prefix_len * self._len))
return (9 # 'chkleaf:\n'
+ len(str(self._maximum_size)) + 1
+ len(str(self._key_width)) + 1
+ len(str(self._len)) + 1
+ prefix_len + 1
+ bytes_for_items)
@classmethod
def deserialise(klass, bytes, key, search_key_func=None):
"""Deserialise bytes, with key key, into a LeafNode.
:param bytes: The bytes of the node.
:param key: The key that the serialised node has.
"""
key = static_tuple.expect_static_tuple(key)
return _deserialise_leaf_node(bytes, key,
search_key_func=search_key_func)
def iteritems(self, store, key_filter=None):
"""Iterate over items in the node.
:param key_filter: A filter to apply to the node. It should be a
list/set/dict or similar repeatedly iterable container.
"""
if key_filter is not None:
# Adjust the filter - short elements go to a prefix filter. All
# other items are looked up directly.
# XXX: perhaps defaultdict? Profiling<rinse and repeat>
filters = {}
for key in key_filter:
if len(key) == self._key_width:
# This filter is meant to match exactly one key, yield it
# if we have it.
try:
yield key, self._items[key]
except KeyError:
# This key is not present in this map, continue
pass
else:
# Short items, we need to match based on a prefix
length_filter = filters.setdefault(len(key), set())
length_filter.add(key)
if filters:
filters = filters.items()
for item in self._items.iteritems():
for length, length_filter in filters:
if item[0][:length] in length_filter:
yield item
break
else:
for item in self._items.iteritems():
yield item
def _key_value_len(self, key, value):
# TODO: Should probably be done without actually joining the key, but
# then that can be done via the C extension
return (len(self._serialise_key(key)) + 1
+ len(str(value.count('\n'))) + 1
+ len(value) + 1)
def _search_key(self, key):
return self._search_key_func(key)
def _map_no_split(self, key, value):
"""Map a key to a value.
This assumes either the key does not already exist, or you have already
removed its size and length from self.
:return: True if adding this node should cause us to split.
"""
self._items[key] = value
self._raw_size += self._key_value_len(key, value)
self._len += 1
serialised_key = self._serialise_key(key)
if self._common_serialised_prefix is None:
self._common_serialised_prefix = serialised_key
else:
self._common_serialised_prefix = self.common_prefix(
self._common_serialised_prefix, serialised_key)
search_key = self._search_key(key)
if self._search_prefix is _unknown:
self._compute_search_prefix()
if self._search_prefix is None:
self._search_prefix = search_key
else:
self._search_prefix = self.common_prefix(
self._search_prefix, search_key)
if (self._len > 1
and self._maximum_size
and self._current_size() > self._maximum_size):
# Check to see if all of the search_keys for this node are
# identical. We allow the node to grow under that circumstance
# (we could track this as common state, but it is infrequent)
if (search_key != self._search_prefix
or not self._are_search_keys_identical()):
return True
return False
def _split(self, store):
"""We have overflowed.
Split this node into multiple LeafNodes, return it up the stack so that
the next layer creates a new InternalNode and references the new nodes.
:return: (common_serialised_prefix, [(node_serialised_prefix, node)])
"""
if self._search_prefix is _unknown:
raise AssertionError('Search prefix must be known')
common_prefix = self._search_prefix
split_at = len(common_prefix) + 1
result = {}
for key, value in self._items.iteritems():
search_key = self._search_key(key)
prefix = search_key[:split_at]
# TODO: Generally only 1 key can be exactly the right length,
# which means we can only have 1 key in the node pointed
# at by the 'prefix\0' key. We might want to consider
# folding it into the containing InternalNode rather than
# having a fixed length-1 node.
# Note this is probably not true for hash keys, as they
# may get a '\00' node anywhere, but won't have keys of
# different lengths.
if len(prefix) < split_at:
prefix += '\x00'*(split_at - len(prefix))
if prefix not in result:
node = LeafNode(search_key_func=self._search_key_func)
node.set_maximum_size(self._maximum_size)
node._key_width = self._key_width
result[prefix] = node
else:
node = result[prefix]
sub_prefix, node_details = node.map(store, key, value)
if len(node_details) > 1:
if prefix != sub_prefix:
# This node has been split and is now found via a different
# path
result.pop(prefix)
new_node = InternalNode(sub_prefix,
search_key_func=self._search_key_func)
new_node.set_maximum_size(self._maximum_size)
new_node._key_width = self._key_width
for split, node in node_details:
new_node.add_node(split, node)
result[prefix] = new_node
return common_prefix, result.items()
def map(self, store, key, value):
"""Map key to value."""
if key in self._items:
self._raw_size -= self._key_value_len(key, self._items[key])
self._len -= 1
self._key = None
if self._map_no_split(key, value):
return self._split(store)
else:
if self._search_prefix is _unknown:
raise AssertionError('%r must be known' % self._search_prefix)
return self._search_prefix, [("", self)]
_serialise_key = '\x00'.join
def serialise(self, store):
"""Serialise the LeafNode to store.
:param store: A VersionedFiles honouring the CHK extensions.
:return: An iterable of the keys inserted by this operation.
"""
lines = ["chkleaf:\n"]
lines.append("%d\n" % self._maximum_size)
lines.append("%d\n" % self._key_width)
lines.append("%d\n" % self._len)
if self._common_serialised_prefix is None:
lines.append('\n')
if len(self._items) != 0:
raise AssertionError('If _common_serialised_prefix is None'
' we should have no items')
else:
lines.append('%s\n' % (self._common_serialised_prefix,))
prefix_len = len(self._common_serialised_prefix)
for key, value in sorted(self._items.items()):
# Always add a final newline
value_lines = osutils.chunks_to_lines([value + '\n'])
serialized = "%s\x00%s\n" % (self._serialise_key(key),
len(value_lines))
if not serialized.startswith(self._common_serialised_prefix):
raise AssertionError('We thought the common prefix was %r'
' but entry %r does not have it in common'
% (self._common_serialised_prefix, serialized))
lines.append(serialized[prefix_len:])
lines.extend(value_lines)
sha1, _, _ = store.add_lines((None,), (), lines)
self._key = StaticTuple("sha1:" + sha1,).intern()
bytes = ''.join(lines)
if len(bytes) != self._current_size():
raise AssertionError('Invalid _current_size')
_get_cache()[self._key] = bytes
return [self._key]
def refs(self):
"""Return the references to other CHK's held by this node."""
return []
def _compute_search_prefix(self):
"""Determine the common search prefix for all keys in this node.
:return: A bytestring of the longest search key prefix that is
unique within this node.
"""
search_keys = [self._search_key_func(key) for key in self._items]
self._search_prefix = self.common_prefix_for_keys(search_keys)
return self._search_prefix
def _are_search_keys_identical(self):
"""Check to see if the search keys for all entries are the same.
When using a hash as the search_key it is possible for non-identical
keys to collide. If that happens enough, we may try overflow a
LeafNode, but as all are collisions, we must not split.
"""
common_search_key = None
for key in self._items:
search_key = self._search_key(key)
if common_search_key is None:
common_search_key = search_key
elif search_key != common_search_key:
return False
return True
def _compute_serialised_prefix(self):
"""Determine the common prefix for serialised keys in this node.
:return: A bytestring of the longest serialised key prefix that is
unique within this node.
"""
serialised_keys = [self._serialise_key(key) for key in self._items]
self._common_serialised_prefix = self.common_prefix_for_keys(
serialised_keys)
return self._common_serialised_prefix
def unmap(self, store, key):
"""Unmap key from the node."""
try:
self._raw_size -= self._key_value_len(key, self._items[key])
except KeyError:
trace.mutter("key %s not found in %r", key, self._items)
raise
self._len -= 1
del self._items[key]
self._key = None
# Recompute from scratch
self._compute_search_prefix()
self._compute_serialised_prefix()
return self
class InternalNode(Node):
"""A node that contains references to other nodes.
An InternalNode is responsible for mapping search key prefixes to child
nodes.
:ivar _items: serialised_key => node dictionary. node may be a tuple,
LeafNode or InternalNode.
"""
__slots__ = ('_node_width',)
def __init__(self, prefix='', search_key_func=None):
Node.__init__(self)
# The size of an internalnode with default values and no children.
# How many octets key prefixes within this node are.
self._node_width = 0
self._search_prefix = prefix
if search_key_func is None:
self._search_key_func = _search_key_plain
else:
self._search_key_func = search_key_func
def add_node(self, prefix, node):
"""Add a child node with prefix prefix, and node node.
:param prefix: The search key prefix for node.
:param node: The node being added.
"""
if self._search_prefix is None:
raise AssertionError("_search_prefix should not be None")
if not prefix.startswith(self._search_prefix):
raise AssertionError("prefixes mismatch: %s must start with %s"
% (prefix,self._search_prefix))
if len(prefix) != len(self._search_prefix) + 1:
raise AssertionError("prefix wrong length: len(%s) is not %d" %
(prefix, len(self._search_prefix) + 1))
self._len += len(node)
if not len(self._items):
self._node_width = len(prefix)
if self._node_width != len(self._search_prefix) + 1:
raise AssertionError("node width mismatch: %d is not %d" %
(self._node_width, len(self._search_prefix) + 1))
self._items[prefix] = node
self._key = None
def _current_size(self):
"""Answer the current serialised size of this node."""
return (self._raw_size + len(str(self._len)) + len(str(self._key_width)) +
len(str(self._maximum_size)))
@classmethod
def deserialise(klass, bytes, key, search_key_func=None):
"""Deserialise bytes to an InternalNode, with key key.
:param bytes: The bytes of the node.
:param key: The key that the serialised node has.
:return: An InternalNode instance.
"""
key = static_tuple.expect_static_tuple(key)
return _deserialise_internal_node(bytes, key,
search_key_func=search_key_func)
def iteritems(self, store, key_filter=None):
for node, node_filter in self._iter_nodes(store, key_filter=key_filter):
for item in node.iteritems(store, key_filter=node_filter):
yield item
def _iter_nodes(self, store, key_filter=None, batch_size=None):
"""Iterate over node objects which match key_filter.
:param store: A store to use for accessing content.
:param key_filter: A key filter to filter nodes. Only nodes that might
contain a key in key_filter will be returned.
:param batch_size: If not None, then we will return the nodes that had
to be read using get_record_stream in batches, rather than reading
them all at once.
:return: An iterable of nodes. This function does not have to be fully
consumed. (There will be no pending I/O when items are being returned.)
"""
# Map from chk key ('sha1:...',) to (prefix, key_filter)
# prefix is the key in self._items to use, key_filter is the key_filter
# entries that would match this node
keys = {}
shortcut = False
if key_filter is None:
# yielding all nodes, yield whatever we have, and queue up a read
# for whatever we are missing
shortcut = True
for prefix, node in self._items.iteritems():
if node.__class__ is StaticTuple:
keys[node] = (prefix, None)
else:
yield node, None
elif len(key_filter) == 1:
# Technically, this path could also be handled by the first check
# in 'self._node_width' in length_filters. However, we can handle
# this case without spending any time building up the
# prefix_to_keys, etc state.
# This is a bit ugly, but TIMEIT showed it to be by far the fastest
# 0.626us list(key_filter)[0]
# is a func() for list(), 2 mallocs, and a getitem
# 0.489us [k for k in key_filter][0]
# still has the mallocs, avoids the func() call
# 0.350us iter(key_filter).next()
# has a func() call, and mallocs an iterator
# 0.125us for key in key_filter: pass
# no func() overhead, might malloc an iterator
# 0.105us for key in key_filter: break
# no func() overhead, might malloc an iterator, probably
# avoids checking an 'else' clause as part of the for
for key in key_filter:
break
search_prefix = self._search_prefix_filter(key)
if len(search_prefix) == self._node_width:
# This item will match exactly, so just do a dict lookup, and
# see what we can return
shortcut = True
try:
node = self._items[search_prefix]
except KeyError:
# A given key can only match 1 child node, if it isn't
# there, then we can just return nothing
return
if node.__class__ is StaticTuple:
keys[node] = (search_prefix, [key])
else:
# This is loaded, and the only thing that can match,
# return
yield node, [key]
return
if not shortcut:
# First, convert all keys into a list of search prefixes
# Aggregate common prefixes, and track the keys they come from
prefix_to_keys = {}
length_filters = {}
for key in key_filter:
search_prefix = self._search_prefix_filter(key)
length_filter = length_filters.setdefault(
len(search_prefix), set())
length_filter.add(search_prefix)
prefix_to_keys.setdefault(search_prefix, []).append(key)
if (self._node_width in length_filters
and len(length_filters) == 1):
# all of the search prefixes match exactly _node_width. This
# means that everything is an exact match, and we can do a
# lookup into self._items, rather than iterating over the items
# dict.
search_prefixes = length_filters[self._node_width]
for search_prefix in search_prefixes:
try:
node = self._items[search_prefix]
except KeyError:
# We can ignore this one
continue
node_key_filter = prefix_to_keys[search_prefix]
if node.__class__ is StaticTuple:
keys[node] = (search_prefix, node_key_filter)
else:
yield node, node_key_filter
else:
# The slow way. We walk every item in self._items, and check to
# see if there are any matches
length_filters = length_filters.items()
for prefix, node in self._items.iteritems():
node_key_filter = []
for length, length_filter in length_filters:
sub_prefix = prefix[:length]
if sub_prefix in length_filter:
node_key_filter.extend(prefix_to_keys[sub_prefix])
if node_key_filter: # this key matched something, yield it
if node.__class__ is StaticTuple:
keys[node] = (prefix, node_key_filter)
else:
yield node, node_key_filter
if keys:
# Look in the page cache for some more bytes
found_keys = set()
for key in keys:
try:
bytes = _get_cache()[key]
except KeyError:
continue
else:
node = _deserialise(bytes, key,
search_key_func=self._search_key_func)
prefix, node_key_filter = keys[key]
self._items[prefix] = node
found_keys.add(key)
yield node, node_key_filter
for key in found_keys:
del keys[key]
if keys:
# demand load some pages.
if batch_size is None:
# Read all the keys in
batch_size = len(keys)
key_order = list(keys)
for batch_start in range(0, len(key_order), batch_size):
batch = key_order[batch_start:batch_start + batch_size]
# We have to fully consume the stream so there is no pending
# I/O, so we buffer the nodes for now.
stream = store.get_record_stream(batch, 'unordered', True)
node_and_filters = []
for record in stream:
bytes = record.get_bytes_as('fulltext')
node = _deserialise(bytes, record.key,
search_key_func=self._search_key_func)
prefix, node_key_filter = keys[record.key]
node_and_filters.append((node, node_key_filter))
self._items[prefix] = node
_get_cache()[record.key] = bytes
for info in node_and_filters:
yield info
def map(self, store, key, value):
"""Map key to value."""
if not len(self._items):
raise AssertionError("can't map in an empty InternalNode.")
search_key = self._search_key(key)
if self._node_width != len(self._search_prefix) + 1:
raise AssertionError("node width mismatch: %d is not %d" %
(self._node_width, len(self._search_prefix) + 1))
if not search_key.startswith(self._search_prefix):
# This key doesn't fit in this index, so we need to split at the
# point where it would fit, insert self into that internal node,
# and then map this key into that node.
new_prefix = self.common_prefix(self._search_prefix,
search_key)
new_parent = InternalNode(new_prefix,
search_key_func=self._search_key_func)
new_parent.set_maximum_size(self._maximum_size)
new_parent._key_width = self._key_width
new_parent.add_node(self._search_prefix[:len(new_prefix)+1],
self)
return new_parent.map(store, key, value)
children = [node for node, _
in self._iter_nodes(store, key_filter=[key])]
if children:
child = children[0]
else:
# new child needed:
child = self._new_child(search_key, LeafNode)
old_len = len(child)
if type(child) is LeafNode:
old_size = child._current_size()
else:
old_size = None
prefix, node_details = child.map(store, key, value)
if len(node_details) == 1:
# child may have shrunk, or might be a new node
child = node_details[0][1]
self._len = self._len - old_len + len(child)
self._items[search_key] = child
self._key = None
new_node = self
if type(child) is LeafNode:
if old_size is None:
# The old node was an InternalNode which means it has now
# collapsed, so we need to check if it will chain to a
# collapse at this level.
trace.mutter("checking remap as InternalNode -> LeafNode")
new_node = self._check_remap(store)
else:
# If the LeafNode has shrunk in size, we may want to run
# a remap check. Checking for a remap is expensive though
# and the frequency of a successful remap is very low.
# Shrinkage by small amounts is common, so we only do the
# remap check if the new_size is low or the shrinkage
# amount is over a configurable limit.
new_size = child._current_size()
shrinkage = old_size - new_size
if (shrinkage > 0 and new_size < _INTERESTING_NEW_SIZE
or shrinkage > _INTERESTING_SHRINKAGE_LIMIT):
trace.mutter(
"checking remap as size shrunk by %d to be %d",
shrinkage, new_size)
new_node = self._check_remap(store)
if new_node._search_prefix is None:
raise AssertionError("_search_prefix should not be None")
return new_node._search_prefix, [('', new_node)]
# child has overflown - create a new intermediate node.
# XXX: This is where we might want to try and expand our depth
# to refer to more bytes of every child (which would give us
# multiple pointers to child nodes, but less intermediate nodes)
child = self._new_child(search_key, InternalNode)
child._search_prefix = prefix
for split, node in node_details:
child.add_node(split, node)
self._len = self._len - old_len + len(child)
self._key = None
return self._search_prefix, [("", self)]
def _new_child(self, search_key, klass):
"""Create a new child node of type klass."""
child = klass()
child.set_maximum_size(self._maximum_size)
child._key_width = self._key_width
child._search_key_func = self._search_key_func
self._items[search_key] = child
return child
def serialise(self, store):
"""Serialise the node to store.
:param store: A VersionedFiles honouring the CHK extensions.
:return: An iterable of the keys inserted by this operation.
"""
for node in self._items.itervalues():
if type(node) is StaticTuple:
# Never deserialised.
continue
if node._key is not None:
# Never altered
continue
for key in node.serialise(store):
yield key
lines = ["chknode:\n"]
lines.append("%d\n" % self._maximum_size)
lines.append("%d\n" % self._key_width)
lines.append("%d\n" % self._len)
if self._search_prefix is None:
raise AssertionError("_search_prefix should not be None")
lines.append('%s\n' % (self._search_prefix,))
prefix_len = len(self._search_prefix)
for prefix, node in sorted(self._items.items()):
if type(node) is StaticTuple:
key = node[0]
else:
key = node._key[0]
serialised = "%s\x00%s\n" % (prefix, key)
if not serialised.startswith(self._search_prefix):
raise AssertionError("prefixes mismatch: %s must start with %s"
% (serialised, self._search_prefix))
lines.append(serialised[prefix_len:])
sha1, _, _ = store.add_lines((None,), (), lines)
self._key = StaticTuple("sha1:" + sha1,).intern()
_get_cache()[self._key] = ''.join(lines)
yield self._key
def _search_key(self, key):
"""Return the serialised key for key in this node."""
# search keys are fixed width. All will be self._node_width wide, so we
# pad as necessary.
return (self._search_key_func(key) + '\x00'*self._node_width)[:self._node_width]
def _search_prefix_filter(self, key):
"""Serialise key for use as a prefix filter in iteritems."""
return self._search_key_func(key)[:self._node_width]
def _split(self, offset):
"""Split this node into smaller nodes starting at offset.
:param offset: The offset to start the new child nodes at.
:return: An iterable of (prefix, node) tuples. prefix is a byte
prefix for reaching node.
"""
if offset >= self._node_width:
for node in self._items.values():
for result in node._split(offset):
yield result
return
for key, node in self._items.items():
pass
def refs(self):
"""Return the references to other CHK's held by this node."""
if self._key is None:
raise AssertionError("unserialised nodes have no refs.")
refs = []
for value in self._items.itervalues():
if type(value) is StaticTuple:
refs.append(value)
else:
refs.append(value.key())
return refs
def _compute_search_prefix(self, extra_key=None):
"""Return the unique key prefix for this node.
:return: A bytestring of the longest search key prefix that is
unique within this node.
"""
self._search_prefix = self.common_prefix_for_keys(self._items)
return self._search_prefix
def unmap(self, store, key, check_remap=True):
"""Remove key from this node and its children."""
if not len(self._items):
raise AssertionError("can't unmap in an empty InternalNode.")
children = [node for node, _
in self._iter_nodes(store, key_filter=[key])]
if children:
child = children[0]
else:
raise KeyError(key)
self._len -= 1
unmapped = child.unmap(store, key)
self._key = None
search_key = self._search_key(key)
if len(unmapped) == 0:
# All child nodes are gone, remove the child:
del self._items[search_key]
unmapped = None
else:
# Stash the returned node
self._items[search_key] = unmapped
if len(self._items) == 1:
# this node is no longer needed:
return self._items.values()[0]
if type(unmapped) is InternalNode:
return self
if check_remap:
return self._check_remap(store)
else:
return self
def _check_remap(self, store):
"""Check if all keys contained by children fit in a single LeafNode.
:param store: A store to use for reading more nodes
:return: Either self, or a new LeafNode which should replace self.
"""
# Logic for how we determine when we need to rebuild
# 1) Implicitly unmap() is removing a key which means that the child
# nodes are going to be shrinking by some extent.
# 2) If all children are LeafNodes, it is possible that they could be
# combined into a single LeafNode, which can then completely replace
# this internal node with a single LeafNode
# 3) If *one* child is an InternalNode, we assume it has already done
# all the work to determine that its children cannot collapse, and
# we can then assume that those nodes *plus* the current nodes don't
# have a chance of collapsing either.
# So a very cheap check is to just say if 'unmapped' is an
# InternalNode, we don't have to check further.
# TODO: Another alternative is to check the total size of all known
# LeafNodes. If there is some formula we can use to determine the
# final size without actually having to read in any more
# children, it would be nice to have. However, we have to be
# careful with stuff like nodes that pull out the common prefix
# of each key, as adding a new key can change the common prefix
# and cause size changes greater than the length of one key.
# So for now, we just add everything to a new Leaf until it
# splits, as we know that will give the right answer
new_leaf = LeafNode(search_key_func=self._search_key_func)
new_leaf.set_maximum_size(self._maximum_size)
new_leaf._key_width = self._key_width
# A batch_size of 16 was chosen because:
# a) In testing, a 4k page held 14 times. So if we have more than 16
# leaf nodes we are unlikely to hold them in a single new leaf
# node. This still allows for 1 round trip
# b) With 16-way fan out, we can still do a single round trip
# c) With 255-way fan out, we don't want to read all 255 and destroy
# the page cache, just to determine that we really don't need it.
for node, _ in self._iter_nodes(store, batch_size=16):
if type(node) is InternalNode:
# Without looking at any leaf nodes, we are sure
return self
for key, value in node._items.iteritems():
if new_leaf._map_no_split(key, value):
return self
trace.mutter("remap generated a new LeafNode")
return new_leaf
def _deserialise(bytes, key, search_key_func):
"""Helper for repositorydetails - convert bytes to a node."""
if bytes.startswith("chkleaf:\n"):
node = LeafNode.deserialise(bytes, key, search_key_func=search_key_func)
elif bytes.startswith("chknode:\n"):
node = InternalNode.deserialise(bytes, key,
search_key_func=search_key_func)
else:
raise AssertionError("Unknown node type.")
return node
class CHKMapDifference(object):
"""Iterate the stored pages and key,value pairs for (new - old).
This class provides a generator over the stored CHK pages and the
(key, value) pairs that are in any of the new maps and not in any of the
old maps.
Note that it may yield chk pages that are common (especially root nodes),
but it won't yield (key,value) pairs that are common.
"""
def __init__(self, store, new_root_keys, old_root_keys,
search_key_func, pb=None):
# TODO: Should we add a StaticTuple barrier here? It would be nice to
# force callers to use StaticTuple, because there will often be
# lots of keys passed in here. And even if we cast it locally,
# that just meanst that we will have *both* a StaticTuple and a
# tuple() in memory, referring to the same object. (so a net
# increase in memory, not a decrease.)
self._store = store
self._new_root_keys = new_root_keys
self._old_root_keys = old_root_keys
self._pb = pb
# All uninteresting chks that we have seen. By the time they are added
# here, they should be either fully ignored, or queued up for
# processing
# TODO: This might grow to a large size if there are lots of merge
# parents, etc. However, it probably doesn't scale to O(history)
# like _processed_new_refs does.
self._all_old_chks = set(self._old_root_keys)
# All items that we have seen from the old_root_keys
self._all_old_items = set()
# These are interesting items which were either read, or already in the
# interesting queue (so we don't need to walk them again)
# TODO: processed_new_refs becomes O(all_chks), consider switching to
# SimpleSet here.
self._processed_new_refs = set()
self._search_key_func = search_key_func
# The uninteresting and interesting nodes to be searched
self._old_queue = []
self._new_queue = []
# Holds the (key, value) items found when processing the root nodes,
# waiting for the uninteresting nodes to be walked
self._new_item_queue = []
self._state = None
def _read_nodes_from_store(self, keys):
# We chose not to use _get_cache(), because we think in
# terms of records to be yielded. Also, we expect to touch each page
# only 1 time during this code. (We may want to evaluate saving the
# raw bytes into the page cache, which would allow a working tree
# update after the fetch to not have to read the bytes again.)
as_st = StaticTuple.from_sequence
stream = self._store.get_record_stream(keys, 'unordered', True)
for record in stream:
if self._pb is not None:
self._pb.tick()
if record.storage_kind == 'absent':
raise errors.NoSuchRevision(self._store, record.key)
bytes = record.get_bytes_as('fulltext')
node = _deserialise(bytes, record.key,
search_key_func=self._search_key_func)
if type(node) is InternalNode:
# Note we don't have to do node.refs() because we know that
# there are no children that have been pushed into this node
# Note: Using as_st() here seemed to save 1.2MB, which would
# indicate that we keep 100k prefix_refs around while
# processing. They *should* be shorter lived than that...
# It does cost us ~10s of processing time
#prefix_refs = [as_st(item) for item in node._items.iteritems()]
prefix_refs = node._items.items()
items = []
else:
prefix_refs = []
# Note: We don't use a StaticTuple here. Profiling showed a
# minor memory improvement (0.8MB out of 335MB peak 0.2%)
# But a significant slowdown (15s / 145s, or 10%)
items = node._items.items()
yield record, node, prefix_refs, items
def _read_old_roots(self):
old_chks_to_enqueue = []
all_old_chks = self._all_old_chks
for record, node, prefix_refs, items in \
self._read_nodes_from_store(self._old_root_keys):
# Uninteresting node
prefix_refs = [p_r for p_r in prefix_refs
if p_r[1] not in all_old_chks]
new_refs = [p_r[1] for p_r in prefix_refs]
all_old_chks.update(new_refs)
# TODO: This might be a good time to turn items into StaticTuple
# instances and possibly intern them. However, this does not
# impact 'initial branch' performance, so I'm not worrying
# about this yet
self._all_old_items.update(items)
# Queue up the uninteresting references
# Don't actually put them in the 'to-read' queue until we have
# finished checking the interesting references
old_chks_to_enqueue.extend(prefix_refs)
return old_chks_to_enqueue
def _enqueue_old(self, new_prefixes, old_chks_to_enqueue):
# At this point, we have read all the uninteresting and interesting
# items, so we can queue up the uninteresting stuff, knowing that we've
# handled the interesting ones
for prefix, ref in old_chks_to_enqueue:
not_interesting = True
for i in xrange(len(prefix), 0, -1):
if prefix[:i] in new_prefixes:
not_interesting = False
break
if not_interesting:
# This prefix is not part of the remaining 'interesting set'
continue
self._old_queue.append(ref)
def _read_all_roots(self):
"""Read the root pages.
This is structured as a generator, so that the root records can be
yielded up to whoever needs them without any buffering.
"""
# This is the bootstrap phase
if not self._old_root_keys:
# With no old_root_keys we can just shortcut and be ready
# for _flush_new_queue
self._new_queue = list(self._new_root_keys)
return
old_chks_to_enqueue = self._read_old_roots()
# filter out any root keys that are already known to be uninteresting
new_keys = set(self._new_root_keys).difference(self._all_old_chks)
# These are prefixes that are present in new_keys that we are
# thinking to yield
new_prefixes = set()
# We are about to yield all of these, so we don't want them getting
# added a second time
processed_new_refs = self._processed_new_refs
processed_new_refs.update(new_keys)
for record, node, prefix_refs, items in \
self._read_nodes_from_store(new_keys):
# At this level, we now know all the uninteresting references
# So we filter and queue up whatever is remaining
prefix_refs = [p_r for p_r in prefix_refs
if p_r[1] not in self._all_old_chks
and p_r[1] not in processed_new_refs]
refs = [p_r[1] for p_r in prefix_refs]
new_prefixes.update([p_r[0] for p_r in prefix_refs])
self._new_queue.extend(refs)
# TODO: We can potentially get multiple items here, however the
# current design allows for this, as callers will do the work
# to make the results unique. We might profile whether we
# gain anything by ensuring unique return values for items
# TODO: This might be a good time to cast to StaticTuple, as
# self._new_item_queue will hold the contents of multiple
# records for an extended lifetime
new_items = [item for item in items
if item not in self._all_old_items]
self._new_item_queue.extend(new_items)
new_prefixes.update([self._search_key_func(item[0])
for item in new_items])
processed_new_refs.update(refs)
yield record
# For new_prefixes we have the full length prefixes queued up.
# However, we also need possible prefixes. (If we have a known ref to
# 'ab', then we also need to include 'a'.) So expand the
# new_prefixes to include all shorter prefixes
for prefix in list(new_prefixes):
new_prefixes.update([prefix[:i] for i in xrange(1, len(prefix))])
self._enqueue_old(new_prefixes, old_chks_to_enqueue)
def _flush_new_queue(self):
# No need to maintain the heap invariant anymore, just pull things out
# and process them
refs = set(self._new_queue)
self._new_queue = []
# First pass, flush all interesting items and convert to using direct refs
all_old_chks = self._all_old_chks
processed_new_refs = self._processed_new_refs
all_old_items = self._all_old_items
new_items = [item for item in self._new_item_queue
if item not in all_old_items]
self._new_item_queue = []
if new_items:
yield None, new_items
refs = refs.difference(all_old_chks)
processed_new_refs.update(refs)
while refs:
# TODO: Using a SimpleSet for self._processed_new_refs and
# saved as much as 10MB of peak memory. However, it requires
# implementing a non-pyrex version.
next_refs = set()
next_refs_update = next_refs.update
# Inlining _read_nodes_from_store improves 'bzr branch bzr.dev'
# from 1m54s to 1m51s. Consider it.
for record, _, p_refs, items in self._read_nodes_from_store(refs):
if all_old_items:
# using the 'if' check saves about 145s => 141s, when
# streaming initial branch of Launchpad data.
items = [item for item in items
if item not in all_old_items]
yield record, items
next_refs_update([p_r[1] for p_r in p_refs])
del p_refs
# set1.difference(set/dict) walks all of set1, and checks if it
# exists in 'other'.
# set1.difference(iterable) walks all of iterable, and does a
# 'difference_update' on a clone of set1. Pick wisely based on the
# expected sizes of objects.
# in our case it is expected that 'new_refs' will always be quite
# small.
next_refs = next_refs.difference(all_old_chks)
next_refs = next_refs.difference(processed_new_refs)
processed_new_refs.update(next_refs)
refs = next_refs
def _process_next_old(self):
# Since we don't filter uninteresting any further than during
# _read_all_roots, process the whole queue in a single pass.
refs = self._old_queue
self._old_queue = []
all_old_chks = self._all_old_chks
for record, _, prefix_refs, items in self._read_nodes_from_store(refs):
# TODO: Use StaticTuple here?
self._all_old_items.update(items)
refs = [r for _,r in prefix_refs if r not in all_old_chks]
self._old_queue.extend(refs)
all_old_chks.update(refs)
def _process_queues(self):
while self._old_queue:
self._process_next_old()
return self._flush_new_queue()
def process(self):
for record in self._read_all_roots():
yield record, []
for record, items in self._process_queues():
yield record, items
def iter_interesting_nodes(store, interesting_root_keys,
uninteresting_root_keys, pb=None):
"""Given root keys, find interesting nodes.
Evaluate nodes referenced by interesting_root_keys. Ones that are also
referenced from uninteresting_root_keys are not considered interesting.
:param interesting_root_keys: keys which should be part of the
"interesting" nodes (which will be yielded)
:param uninteresting_root_keys: keys which should be filtered out of the
result set.
:return: Yield
(interesting record, {interesting key:values})
"""
iterator = CHKMapDifference(store, interesting_root_keys,
uninteresting_root_keys,
search_key_func=store._search_key_func,
pb=pb)
return iterator.process()
try:
from bzrlib._chk_map_pyx import (
_bytes_to_text_key,
_search_key_16,
_search_key_255,
_deserialise_leaf_node,
_deserialise_internal_node,
)
except ImportError, e:
osutils.failed_to_load_extension(e)
from bzrlib._chk_map_py import (
_bytes_to_text_key,
_search_key_16,
_search_key_255,
_deserialise_leaf_node,
_deserialise_internal_node,
)
search_key_registry.register('hash-16-way', _search_key_16)
search_key_registry.register('hash-255-way', _search_key_255)
def _check_key(key):
"""Helper function to assert that a key is properly formatted.
This generally shouldn't be used in production code, but it can be helpful
to debug problems.
"""
if type(key) is not StaticTuple:
raise TypeError('key %r is not StaticTuple but %s' % (key, type(key)))
if len(key) != 1:
raise ValueError('key %r should have length 1, not %d' % (key, len(key),))
if type(key[0]) is not str:
raise TypeError('key %r should hold a str, not %r'
% (key, type(key[0])))
if not key[0].startswith('sha1:'):
raise ValueError('key %r should point to a sha1:' % (key,))
|