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 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174
// Copyright Materialize, Inc. and contributors. All rights reserved.
//
// Use of this software is governed by the Business Source License
// included in the LICENSE file.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0.
#![warn(missing_docs)]
use std::cmp::Ordering;
use std::collections::HashSet;
use std::fmt;
use itertools::Itertools;
use serde::{Deserialize, Serialize};
use lowertest::MzReflect;
use ore::collections::CollectionExt;
use ore::id_gen::IdGen;
use ore::stack::{maybe_grow, CheckedRecursion, RecursionGuard, RecursionLimitError};
use repr::{ColumnName, ColumnType, Datum, Diff, RelationType, Row, ScalarType};
use self::func::{AggregateFunc, TableFunc};
use crate::explain::ViewExplanation;
use crate::{
func as scalar_func, BinaryFunc, DummyHumanizer, EvalError, ExprHumanizer, GlobalId, Id,
LocalId, MirScalarExpr, UnaryFunc, VariadicFunc,
};
pub mod canonicalize;
pub mod func;
pub mod join_input_mapper;
/// A recursion limit to be used for stack-safe traversals of [`MirRelationExpr`] trees.
///
/// The recursion limit must be large enough to accomodate for the linear representation
/// of some pathological but frequently occurring query fragments.
///
/// For example, in MIR we could have long chains of
/// - (1) `Let` bindings,
/// - (2) `CallBinary` calls with associative functions such as `OR` and `+`
///
/// Until we fix those, we need to stick with the larger recursion limit.
pub const RECURSION_LIMIT: usize = 2048;
/// An abstract syntax tree which defines a collection.
///
/// The AST is meant reflect the capabilities of the `differential_dataflow::Collection` type,
/// written generically enough to avoid run-time compilation work.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize, Hash, MzReflect)]
pub enum MirRelationExpr {
/// A constant relation containing specified rows.
///
/// The runtime memory footprint of this operator is zero.
Constant {
/// Rows of the constant collection and their multiplicities.
rows: Result<Vec<(Row, Diff)>, EvalError>,
/// Schema of the collection.
typ: RelationType,
},
/// Get an existing dataflow.
///
/// The runtime memory footprint of this operator is zero.
Get {
/// The identifier for the collection to load.
#[mzreflect(ignore)]
id: Id,
/// Schema of the collection.
typ: RelationType,
},
/// Introduce a temporary dataflow.
///
/// The runtime memory footprint of this operator is zero.
Let {
/// The identifier to be used in `Get` variants to retrieve `value`.
#[mzreflect(ignore)]
id: LocalId,
/// The collection to be bound to `id`.
value: Box<MirRelationExpr>,
/// The result of the `Let`, evaluated with `id` bound to `value`.
body: Box<MirRelationExpr>,
},
/// Project out some columns from a dataflow
///
/// The runtime memory footprint of this operator is zero.
Project {
/// The source collection.
input: Box<MirRelationExpr>,
/// Indices of columns to retain.
outputs: Vec<usize>,
},
/// Append new columns to a dataflow
///
/// The runtime memory footprint of this operator is zero.
Map {
/// The source collection.
input: Box<MirRelationExpr>,
/// Expressions which determine values to append to each row.
/// An expression may refer to columns in `input` or
/// expressions defined earlier in the vector
scalars: Vec<MirScalarExpr>,
},
/// Like Map, but yields zero-or-more output rows per input row
///
/// The runtime memory footprint of this operator is zero.
FlatMap {
/// The source collection
input: Box<MirRelationExpr>,
/// The table func to apply
func: TableFunc,
/// The argument to the table func
exprs: Vec<MirScalarExpr>,
},
/// Keep rows from a dataflow where all the predicates are true
///
/// The runtime memory footprint of this operator is zero.
Filter {
/// The source collection.
input: Box<MirRelationExpr>,
/// Predicates, each of which must be true.
predicates: Vec<MirScalarExpr>,
},
/// Join several collections, where some columns must be equal.
///
/// For further details consult the documentation for [`MirRelationExpr::join`].
///
/// The runtime memory footprint of this operator can be proportional to
/// the sizes of all inputs and the size of all joins of prefixes.
/// This may be reduced due to arrangements available at rendering time.
Join {
/// A sequence of input relations.
inputs: Vec<MirRelationExpr>,
/// A sequence of equivalence classes of expressions on the cross product of inputs.
///
/// Each equivalence class is a list of scalar expressions, where for each class the
/// intended interpretation is that all evaluated expressions should be equal.
///
/// Each scalar expression is to be evaluated over the cross-product of all records
/// from all inputs. In many cases this may just be column selection from specific
/// inputs, but more general cases exist (e.g. complex functions of multiple columns
/// from multiple inputs, or just constant literals).
equivalences: Vec<Vec<MirScalarExpr>>,
/// Join implementation information.
#[serde(default)]
implementation: JoinImplementation,
},
/// Group a dataflow by some columns and aggregate over each group
///
/// The runtime memory footprint of this operator is at most proportional to the
/// number of distinct records in the input and output. The actual requirements
/// can be less: the number of distinct inputs to each aggregate, summed across
/// each aggregate, plus the output size. For more details consult the code that
/// builds the associated dataflow.
Reduce {
/// The source collection.
input: Box<MirRelationExpr>,
/// Column indices used to form groups.
group_key: Vec<MirScalarExpr>,
/// Expressions which determine values to append to each row, after the group keys.
aggregates: Vec<AggregateExpr>,
/// True iff the input is known to monotonically increase (only addition of records).
#[serde(default)]
monotonic: bool,
/// User hint: expected number of values per group key. Used to optimize physical rendering.
#[serde(default)]
expected_group_size: Option<usize>,
},
/// Groups and orders within each group, limiting output.
///
/// The runtime memory footprint of this operator is proportional to its input and output.
TopK {
/// The source collection.
input: Box<MirRelationExpr>,
/// Column indices used to form groups.
group_key: Vec<usize>,
/// Column indices used to order rows within groups.
order_key: Vec<ColumnOrder>,
/// Number of records to retain
#[serde(default)]
limit: Option<usize>,
/// Number of records to skip
#[serde(default)]
offset: usize,
/// True iff the input is known to monotonically increase (only addition of records).
#[serde(default)]
monotonic: bool,
},
/// Return a dataflow where the row counts are negated
///
/// The runtime memory footprint of this operator is zero.
Negate {
/// The source collection.
input: Box<MirRelationExpr>,
},
/// Keep rows from a dataflow where the row counts are positive
///
/// The runtime memory footprint of this operator is proportional to its input and output.
Threshold {
/// The source collection.
input: Box<MirRelationExpr>,
},
/// Adds the frequencies of elements in contained sets.
///
/// The runtime memory footprint of this operator is zero.
Union {
/// A source collection.
base: Box<MirRelationExpr>,
/// Source collections to union.
inputs: Vec<MirRelationExpr>,
},
/// Technically a no-op. Used to render an index. Will be used to optimize queries
/// on finer grain
///
/// The runtime memory footprint of this operator is proportional to its input.
ArrangeBy {
/// The source collection
input: Box<MirRelationExpr>,
/// Columns to arrange `input` by, in order of decreasing primacy
keys: Vec<Vec<MirScalarExpr>>,
},
/// Declares that `keys` are primary keys for `input`.
/// Should be used *very* sparingly, and only if there's no plausible
/// way to derive the key information from the underlying expression.
/// The result of declaring a key that isn't actually a key for the underlying expression is undefined.
///
/// There is no operator rendered for this IR node; thus, its runtime memory footprint is zero.
DeclareKeys {
/// The source collection
input: Box<MirRelationExpr>,
/// The set of columns in the source collection that form a key.
keys: Vec<Vec<usize>>,
},
}
impl MirRelationExpr {
/// Reports the schema of the relation.
///
/// This method determines the type through recursive traversal of the
/// relation expression, drawing from the types of base collections.
/// As such, this is not an especially cheap method, and should be used
/// judiciously.
///
/// The relation type is computed incrementally with a recursive post-order
/// traversal, that accumulates the input types for the relations yet to be
/// visited in `type_stack`.
pub fn typ(&self) -> RelationType {
let mut type_stack = Vec::new();
self.visit_pre_post(
&mut |e: &MirRelationExpr| -> Option<Vec<&MirRelationExpr>> {
if let MirRelationExpr::Let { body, .. } = &e {
// Do not traverse the value sub-graph, since it's not relevant for
// determing the relation type of Let operators.
Some(vec![&*body])
} else {
None
}
},
&mut |e: &MirRelationExpr| {
if let MirRelationExpr::Let { .. } = &e {
let body_typ = type_stack.pop().unwrap();
// Insert a dummy relation type for the value, since `typ_with_input_types`
// won't look at it, but expects the relation type of the body to be second.
type_stack.push(RelationType::empty());
type_stack.push(body_typ);
}
let num_inputs = e.num_inputs();
let relation_type =
e.typ_with_input_types(&type_stack[type_stack.len() - num_inputs..]);
type_stack.truncate(type_stack.len() - num_inputs);
type_stack.push(relation_type);
},
);
assert_eq!(type_stack.len(), 1);
type_stack.pop().unwrap()
}
/// Reports the schema of the relation given the schema of the input relations.
///
/// `input_types` is required to contain the schemas for the input relations of
/// the current relation in the same order as they are visited by `try_visit1`
/// method, even though not all may be used for computing the schema of the
/// current relation. For example, `Let` expects two input types, one for the
/// value relation and one for the body, in that order, but only the one for the
/// body is used to determine the type of the `Let` relation.
///
/// It is meant to be used during post-order traversals to compute relation
/// schemas incrementally.
pub fn typ_with_input_types(&self, input_types: &[RelationType]) -> RelationType {
assert_eq!(self.num_inputs(), input_types.len());
match self {
MirRelationExpr::Constant { rows, typ } => {
if let Ok(rows) = rows {
let n_cols = typ.arity();
// If the `i`th entry is `Some`, then we have not yet observed non-uniqueness in the `i`th column.
let mut unique_values_per_col = vec![Some(HashSet::<Datum>::default()); n_cols];
for (row, diff) in rows {
for (i, (datum, column_typ)) in
row.iter().zip(typ.column_types.iter()).enumerate()
{
// If the record will be observed, we should validate its type.
if datum != Datum::Dummy {
assert!(
datum.is_instance_of(column_typ),
"Expected datum of type {:?}, got value {:?}",
column_typ,
datum
);
if let Some(unique_vals) = &mut unique_values_per_col[i] {
let is_dupe = *diff != 1 || !unique_vals.insert(datum);
if is_dupe {
unique_values_per_col[i] = None;
}
}
}
}
}
if rows.len() == 0 || (rows.len() == 1 && rows[0].1 == 1) {
RelationType::new(typ.column_types.clone()).with_key(vec![])
} else {
// XXX - Multi-column keys are not detected.
typ.clone().with_keys(
unique_values_per_col
.into_iter()
.enumerate()
.filter(|(_idx, unique_vals)| unique_vals.is_some())
.map(|(idx, _)| vec![idx])
.collect(),
)
}
} else {
typ.clone()
}
}
MirRelationExpr::Get { typ, .. } => typ.clone(),
MirRelationExpr::Let { .. } => input_types.last().unwrap().clone(),
MirRelationExpr::Project { input: _, outputs } => {
let input_typ = &input_types[0];
let mut output_typ = RelationType::new(
outputs
.iter()
.map(|&i| input_typ.column_types[i].clone())
.collect(),
);
for keys in input_typ.keys.iter() {
if keys.iter().all(|k| outputs.contains(k)) {
output_typ = output_typ.with_key(
keys.iter()
.map(|c| outputs.iter().position(|o| o == c).unwrap())
.collect(),
);
}
}
output_typ
}
MirRelationExpr::Map { scalars, .. } => {
let mut typ = input_types[0].clone();
let arity = typ.column_types.len();
let mut remappings = Vec::new();
for (column, scalar) in scalars.iter().enumerate() {
typ.column_types.push(scalar.typ(&typ));
// assess whether the scalar preserves uniqueness,
// and could participate in a key!
fn uniqueness(expr: &MirScalarExpr) -> Option<usize> {
match expr {
MirScalarExpr::CallUnary { func, expr } => {
if func.preserves_uniqueness() {
uniqueness(expr)
} else {
None
}
}
MirScalarExpr::Column(c) => Some(*c),
_ => None,
}
}
if let Some(c) = uniqueness(scalar) {
remappings.push((c, column + arity));
}
}
// Any column in `remappings` could be replaced in a key
// by the corresponding c. This could lead to combinatorial
// explosion using our current representation, so we wont
// do that. Instead, we'll handle the case of one remapping.
if remappings.len() == 1 {
let (old, new) = remappings.pop().unwrap();
let mut new_keys = Vec::new();
for key in typ.keys.iter() {
if key.contains(&old) {
let mut new_key: Vec<usize> =
key.iter().cloned().filter(|k| k != &old).collect();
new_key.push(new);
new_key.sort_unstable();
new_keys.push(new_key);
}
}
for new_key in new_keys {
typ = typ.with_key(new_key);
}
}
typ
}
MirRelationExpr::FlatMap { func, .. } => {
let mut input_typ = input_types[0].clone();
input_typ
.column_types
.extend(func.output_type().column_types);
// FlatMap can add duplicate rows, so input keys are no longer valid
let typ = RelationType::new(input_typ.column_types);
typ
}
MirRelationExpr::Filter { predicates, .. } => {
// A filter inherits the keys of its input unless the filters
// have reduced the input to a single row, in which case the
// keys of the input are `()`.
let mut input_typ = input_types[0].clone();
let cols_equal_to_literal = predicates
.iter()
.filter_map(|p| {
if let MirScalarExpr::CallBinary {
func: crate::BinaryFunc::Eq,
expr1,
expr2,
} = p
{
if let MirScalarExpr::Column(c) = &**expr1 {
if expr2.is_literal_ok() {
return Some(c);
}
}
}
None
})
.collect::<Vec<_>>();
for key_set in &mut input_typ.keys {
key_set.retain(|k| !cols_equal_to_literal.contains(&k));
}
if !input_typ.keys.is_empty() {
// If `[0 1]` is an input key and there `#0 = #1` exists as a
// predicate, we should present both `[0]` and `[1]` as keys
// of the output. Also, if there is a key involving X column
// and an equality between X and another column Y, a variant
// of that key with Y instead of X should be presented as
// a key of the output.
// First, we build an iterator over the equivalences
let classes = predicates.iter().filter_map(|p| {
if let MirScalarExpr::CallBinary {
func: crate::BinaryFunc::Eq,
expr1,
expr2,
} = p
{
if let Some(c1) = expr1.as_column() {
if let Some(c2) = expr2.as_column() {
return Some((c1, c2));
}
}
}
None
});
// Keep doing replacements until the number of keys settles
let mut prev_keys: HashSet<_> = input_typ.keys.drain(..).collect();
let mut prev_keys_size = 0;
while prev_keys_size != prev_keys.len() {
prev_keys_size = prev_keys.len();
for (c1, c2) in classes.clone() {
let mut new_keys = HashSet::new();
for key in prev_keys.into_iter() {
let contains_c1 = key.contains(&c1);
let contains_c2 = key.contains(&c2);
if contains_c1 && contains_c2 {
new_keys.insert(
key.iter().filter(|c| **c != c1).cloned().collect_vec(),
);
new_keys.insert(
key.iter().filter(|c| **c != c2).cloned().collect_vec(),
);
} else {
if contains_c1 {
new_keys.insert(
key.iter()
.map(|c| if *c == c1 { c2 } else { *c })
.sorted()
.collect_vec(),
);
} else if contains_c2 {
new_keys.insert(
key.iter()
.map(|c| if *c == c2 { c1 } else { *c })
.sorted()
.collect_vec(),
);
}
new_keys.insert(key);
}
}
prev_keys = new_keys;
}
}
input_typ.keys = prev_keys.into_iter().sorted().collect_vec();
}
// Augment non-nullability of columns, by observing either
// 1. Predicates that explicitly test for null values, and
// 2. Columns that if null would make a predicate be null.
let mut nonnull_required_columns = HashSet::new();
for predicate in predicates {
// Add any columns that being null would force the predicate to be null.
// Should that happen, the row would be discarded.
predicate.non_null_requirements(&mut nonnull_required_columns);
// Test for explicit checks that a column is non-null.
if let MirScalarExpr::CallUnary {
func: UnaryFunc::Not(scalar_func::Not),
expr,
} = predicate
{
if let MirScalarExpr::CallUnary {
func: UnaryFunc::IsNull(scalar_func::IsNull),
expr,
} = &**expr
{
if let MirScalarExpr::Column(c) = &**expr {
input_typ.column_types[*c].nullable = false;
}
}
}
}
// Set as nonnull any columns where null values would cause
// any predicate to evaluate to null.
for column in nonnull_required_columns.into_iter() {
input_typ.column_types[column].nullable = false;
}
input_typ
}
MirRelationExpr::Join { equivalences, .. } => {
// Iterating and cloning types inside the flat_map() avoids allocating Vec<>,
// as clones are directly added to column_types Vec<>.
let column_types = input_types
.iter()
.flat_map(|i| i.column_types.iter().cloned())
.collect::<Vec<_>>();
let mut typ = RelationType::new(column_types);
// It is important the `new_from_input_types` constructor is
// used. Otherwise, Materialize may potentially end up in an
// infinite loop.
let input_mapper =
join_input_mapper::JoinInputMapper::new_from_input_types(input_types);
let global_keys = input_mapper.global_keys(
&input_types
.iter()
.map(|t| t.keys.clone())
.collect::<Vec<_>>(),
equivalences,
);
for keys in global_keys {
typ = typ.with_key(keys.clone());
}
typ
}
MirRelationExpr::Reduce {
group_key,
aggregates,
..
} => {
let input_typ = &input_types[0];
let mut column_types = group_key
.iter()
.map(|e| e.typ(input_typ))
.collect::<Vec<_>>();
for agg in aggregates {
column_types.push(agg.typ(input_typ));
}
let mut result = RelationType::new(column_types);
// The group key should form a key, but we might already have
// keys that are subsets of the group key, and should retain
// those instead, if so.
let mut keys = Vec::new();
for key in input_typ.keys.iter() {
if key
.iter()
.all(|k| group_key.contains(&MirScalarExpr::Column(*k)))
{
keys.push(
key.iter()
.map(|i| {
group_key
.iter()
.position(|k| k == &MirScalarExpr::Column(*i))
.unwrap()
})
.collect::<Vec<_>>(),
);
}
}
if keys.is_empty() {
keys.push((0..group_key.len()).collect());
}
for key in keys {
result = result.with_key(key);
}
result
}
MirRelationExpr::TopK {
group_key, limit, ..
} => {
// If `limit` is `Some(1)` then the group key will become
// a unique key, as there will be only one record with that key.
let mut typ = input_types[0].clone();
if limit == &Some(1) {
typ = typ.with_key(group_key.clone())
}
typ
}
MirRelationExpr::Negate { input: _ } => {
// Although negate may have distinct records for each key,
// the multiplicity is -1 rather than 1. This breaks many
// of the optimization uses of "keys".
let mut typ = input_types[0].clone();
typ.keys.clear();
typ
}
MirRelationExpr::Threshold { .. } => input_types[0].clone(),
MirRelationExpr::Union { base, inputs } => {
let mut base_cols = input_types[0].column_types.clone();
for input_type in input_types.iter().skip(1) {
for (base_col, col) in
base_cols.iter_mut().zip_eq(input_type.column_types.iter())
{
*base_col = base_col
.union(&col)
.map_err(|e| format!("{}\nIn {:#?}", e, self))
.unwrap();
}
}
// Generally, unions do not have any unique keys, because
// each input might duplicate some. However, there is at
// least one idiomatic structure that does preserve keys,
// which results from SQL aggregations that must populate
// absent records with default values. In that pattern,
// the union of one GET with its negation, which has first
// been subjected to a projection and map, we can remove
// their influence on the key structure.
//
// If there are A, B, each with a unique `key` such that
// we are looking at
//
// A.proj(set_containg_key) + (B - A.proj(key)).map(stuff)
//
// Then we can report `key` as a unique key.
//
// TODO: make unique key structure an optimization analysis
// rather than part of the type information.
// TODO: perhaps ensure that (above) A.proj(key) is a
// subset of B, as otherwise there are negative records
// and who knows what is true (not expected, but again
// who knows what the query plan might look like).
let (base_projection, base_with_project_stripped) =
if let MirRelationExpr::Project { input, outputs } = &**base {
(outputs.clone(), &**input)
} else {
// A input without a project is equivalent to an input
// with the project being all columns in the input in order.
((0..base_cols.len()).collect::<Vec<_>>(), &**base)
};
let mut keys = Vec::new();
if let MirRelationExpr::Get {
id: first_id,
typ: _,
} = base_with_project_stripped
{
if inputs.len() == 1 {
if let MirRelationExpr::Map { input, .. } = &inputs[0] {
if let MirRelationExpr::Union { base, inputs } = &**input {
if inputs.len() == 1 {
if let MirRelationExpr::Project { input, outputs } = &**base {
if let MirRelationExpr::Negate { input } = &**input {
if let MirRelationExpr::Get {
id: second_id,
typ: _,
} = &**input
{
if first_id == second_id {
keys.extend(
inputs[0].typ().keys.drain(..).filter(
|key| {
key.iter().all(|c| {
outputs.get(*c) == Some(c)
&& base_projection.get(*c)
== Some(c)
})
},
),
);
}
}
}
}
}
}
}
}
}
RelationType::new(base_cols).with_keys(keys)
// Important: do not inherit keys of either input, as not unique.
}
MirRelationExpr::ArrangeBy { .. } => input_types[0].clone(),
MirRelationExpr::DeclareKeys { keys, .. } => {
input_types[0].clone().with_keys(keys.clone())
}
}
}
/// The number of columns in the relation.
///
/// This number is determined from the type, which is determined recursively
/// at non-trivial cost.
pub fn arity(&self) -> usize {
match self {
MirRelationExpr::Constant { rows: _, typ } => typ.arity(),
MirRelationExpr::Get { typ, .. } => typ.arity(),
MirRelationExpr::Let { body, .. } => body.arity(),
MirRelationExpr::Project { input: _, outputs } => outputs.len(),
MirRelationExpr::Map { input, scalars } => input.arity() + scalars.len(),
MirRelationExpr::FlatMap { input, func, .. } => {
input.arity() + func.output_type().column_types.len()
}
MirRelationExpr::Filter { input, .. } => input.arity(),
MirRelationExpr::Join { inputs, .. } => inputs.iter().map(|i| i.arity()).sum(),
MirRelationExpr::Reduce {
input: _,
group_key,
aggregates,
..
} => group_key.len() + aggregates.len(),
MirRelationExpr::TopK { input, .. } => input.arity(),
MirRelationExpr::Negate { input } => input.arity(),
MirRelationExpr::Threshold { input } => input.arity(),
MirRelationExpr::Union { base, inputs: _ } => base.arity(),
MirRelationExpr::ArrangeBy { input, .. } => input.arity(),
MirRelationExpr::DeclareKeys { input, .. } => input.arity(),
}
}
/// The number of child relations this relation has.
pub fn num_inputs(&self) -> usize {
let mut count = 0;
self.visit_children(|_| count += 1);
count
}
/// Constructs a constant collection from specific rows and schema, where
/// each row will have a multiplicity of one.
pub fn constant(rows: Vec<Vec<Datum>>, typ: RelationType) -> Self {
let rows = rows.into_iter().map(|row| (row, 1)).collect();
MirRelationExpr::constant_diff(rows, typ)
}
/// Constructs a constant collection from specific rows and schema, where
/// each row can have an arbitrary multiplicity.
pub fn constant_diff(rows: Vec<(Vec<Datum>, Diff)>, typ: RelationType) -> Self {
for (row, _diff) in &rows {
for (datum, column_typ) in row.iter().zip(typ.column_types.iter()) {
assert!(
datum.is_instance_of(column_typ),
"Expected datum of type {:?}, got value {:?}",
column_typ,
datum
);
}
}
let rows = Ok(rows
.into_iter()
.map(move |(row, diff)| (Row::pack_slice(&row), diff))
.collect());
MirRelationExpr::Constant { rows, typ }
}
/// Constructs the expression for getting a global collection
pub fn global_get(id: GlobalId, typ: RelationType) -> Self {
MirRelationExpr::Get {
id: Id::Global(id),
typ,
}
}
/// Retains only the columns specified by `output`.
pub fn project(self, outputs: Vec<usize>) -> Self {
MirRelationExpr::Project {
input: Box::new(self),
outputs,
}
}
/// Append to each row the results of applying elements of `scalar`.
pub fn map(self, scalars: Vec<MirScalarExpr>) -> Self {
MirRelationExpr::Map {
input: Box::new(self),
scalars,
}
}
/// Like `map`, but yields zero-or-more output rows per input row
pub fn flat_map(self, func: TableFunc, exprs: Vec<MirScalarExpr>) -> Self {
MirRelationExpr::FlatMap {
input: Box::new(self),
func,
exprs,
}
}
/// Retain only the rows satisifying each of several predicates.
pub fn filter<I>(self, predicates: I) -> Self
where
I: IntoIterator<Item = MirScalarExpr>,
{
MirRelationExpr::Filter {
input: Box::new(self),
predicates: predicates.into_iter().collect(),
}
}
/// Form the Cartesian outer-product of rows in both inputs.
pub fn product(self, right: Self) -> Self {
MirRelationExpr::join(vec![self, right], vec![])
}
/// Performs a relational equijoin among the input collections.
///
/// The sequence `inputs` each describe different input collections, and the sequence `variables` describes
/// equality constraints that some of their columns must satisfy. Each element in `variable` describes a set
/// of pairs `(input_index, column_index)` where every value described by that set must be equal.
///
/// For example, the pair `(input, column)` indexes into `inputs[input][column]`, extracting the `input`th
/// input collection and for each row examining its `column`th column.
///
/// # Example
///
/// ```rust
/// use repr::{Datum, ColumnType, RelationType, ScalarType};
/// use expr::MirRelationExpr;
///
/// // A common schema for each input.
/// let schema = RelationType::new(vec![
/// ScalarType::Int32.nullable(false),
/// ScalarType::Int32.nullable(false),
/// ]);
///
/// // the specific data are not important here.
/// let data = vec![Datum::Int32(0), Datum::Int32(1)];
///
/// // Three collections that could have been different.
/// let input0 = MirRelationExpr::constant(vec![data.clone()], schema.clone());
/// let input1 = MirRelationExpr::constant(vec![data.clone()], schema.clone());
/// let input2 = MirRelationExpr::constant(vec![data.clone()], schema.clone());
///
/// // Join the three relations looking for triangles, like so.
/// //
/// // Output(A,B,C) := Input0(A,B), Input1(B,C), Input2(A,C)
/// let joined = MirRelationExpr::join(
/// vec![input0, input1, input2],
/// vec![
/// vec![(0,0), (2,0)], // fields A of inputs 0 and 2.
/// vec![(0,1), (1,0)], // fields B of inputs 0 and 1.
/// vec![(1,1), (2,1)], // fields C of inputs 1 and 2.
/// ],
/// );
///
/// // Technically the above produces `Output(A,B,B,C,A,C)` because the columns are concatenated.
/// // A projection resolves this and produces the correct output.
/// let result = joined.project(vec![0, 1, 3]);
/// ```
pub fn join(inputs: Vec<MirRelationExpr>, variables: Vec<Vec<(usize, usize)>>) -> Self {
let input_mapper = join_input_mapper::JoinInputMapper::new(&inputs);
let equivalences = variables
.into_iter()
.map(|vs| {
vs.into_iter()
.map(|(r, c)| input_mapper.map_expr_to_global(MirScalarExpr::Column(c), r))
.collect::<Vec<_>>()
})
.collect::<Vec<_>>();
Self::join_scalars(inputs, equivalences)
}
/// Constructs a join operator from inputs and required-equal scalar expressions.
pub fn join_scalars(
inputs: Vec<MirRelationExpr>,
equivalences: Vec<Vec<MirScalarExpr>>,
) -> Self {
MirRelationExpr::Join {
inputs,
equivalences,
implementation: JoinImplementation::Unimplemented,
}
}
/// Perform a key-wise reduction / aggregation.
///
/// The `group_key` argument indicates columns in the input collection that should
/// be grouped, and `aggregates` lists aggregation functions each of which produces
/// one output column in addition to the keys.
pub fn reduce(
self,
group_key: Vec<usize>,
aggregates: Vec<AggregateExpr>,
expected_group_size: Option<usize>,
) -> Self {
MirRelationExpr::Reduce {
input: Box::new(self),
group_key: group_key.into_iter().map(MirScalarExpr::Column).collect(),
aggregates,
monotonic: false,
expected_group_size,
}
}
/// Perform a key-wise reduction order by and limit.
///
/// The `group_key` argument indicates columns in the input collection that should
/// be grouped, the `order_key` argument indicates columns that should be further
/// used to order records within groups, and the `limit` argument constrains the
/// total number of records that should be produced in each group.
pub fn top_k(
self,
group_key: Vec<usize>,
order_key: Vec<ColumnOrder>,
limit: Option<usize>,
offset: usize,
) -> Self {
MirRelationExpr::TopK {
input: Box::new(self),
group_key,
order_key,
limit,
offset,
monotonic: false,
}
}
/// Negates the occurrences of each row.
pub fn negate(self) -> Self {
MirRelationExpr::Negate {
input: Box::new(self),
}
}
/// Removes all but the first occurrence of each row.
pub fn distinct(self) -> Self {
let arity = self.arity();
self.distinct_by((0..arity).collect())
}
/// Removes all but the first occurrence of each key. Columns not included
/// in the `group_key` are discarded.
pub fn distinct_by(self, group_key: Vec<usize>) -> Self {
self.reduce(group_key, vec![], None)
}
/// Discards rows with a negative frequency.
pub fn threshold(self) -> Self {
MirRelationExpr::Threshold {
input: Box::new(self),
}
}
/// Unions together any number inputs.
///
/// If `inputs` is empty, then an empty relation of type `typ` is
/// constructed.
pub fn union_many(mut inputs: Vec<Self>, typ: RelationType) -> Self {
if inputs.len() == 0 {
MirRelationExpr::Constant {
rows: Ok(vec![]),
typ,
}
} else if inputs.len() == 1 {
inputs.into_element()
} else {
MirRelationExpr::Union {
base: Box::new(inputs.remove(0)),
inputs,
}
}
}
/// Produces one collection where each row is present with the sum of its frequencies in each input.
pub fn union(self, other: Self) -> Self {
MirRelationExpr::Union {
base: Box::new(self),
inputs: vec![other],
}
}
/// Arranges the collection by the specified columns
pub fn arrange_by(self, keys: &[Vec<MirScalarExpr>]) -> Self {
MirRelationExpr::ArrangeBy {
input: Box::new(self),
keys: keys.to_owned(),
}
}
/// Indicates if this is a constant empty collection.
///
/// A false value does not mean the collection is known to be non-empty,
/// only that we cannot currently determine that it is statically empty.
pub fn is_empty(&self) -> bool {
if let MirRelationExpr::Constant { rows: Ok(rows), .. } = self {
rows.is_empty()
} else {
false
}
}
/// Returns the distinct global identifiers on which this expression
/// depends.
///
/// See [`MirRelationExpr::global_uses_into`] to reuse an existing vector.
pub fn global_uses(&self) -> Vec<GlobalId> {
let mut out = vec![];
self.global_uses_into(&mut out);
out.sort();
out.dedup();
out
}
/// Appends global identifiers on which this expression depends to `out`.
///
/// Unlike [`MirRelationExpr::global_uses`], this method does not deduplicate
/// the global identifiers.
pub fn global_uses_into(&self, out: &mut Vec<GlobalId>) {
if let MirRelationExpr::Get {
id: Id::Global(id), ..
} = self
{
out.push(*id);
}
self.visit_children(|expr| expr.global_uses_into(out))
}
/// Pretty-print this MirRelationExpr to a string.
///
/// This method allows an additional `ExprHumanizer` which can annotate
/// identifiers with human-meaningful names for the identifiers.
pub fn pretty_humanized(&self, id_humanizer: &impl ExprHumanizer) -> String {
ViewExplanation::new(self, id_humanizer).to_string()
}
/// Pretty-print this MirRelationExpr to a string.
pub fn pretty(&self) -> String {
ViewExplanation::new(self, &DummyHumanizer).to_string()
}
/// Print this MirRelationExpr to a JSON-formatted string.
pub fn json(&self) -> String {
serde_json::to_string(self).unwrap()
}
/// Pretty-print this MirRelationExpr to a string with type information.
pub fn pretty_typed(&self) -> String {
let mut explanation = ViewExplanation::new(self, &DummyHumanizer);
explanation.explain_types();
explanation.to_string()
}
/// Take ownership of `self`, leaving an empty `MirRelationExpr::Constant` with the correct type.
pub fn take_safely(&mut self) -> MirRelationExpr {
let typ = self.typ();
std::mem::replace(
self,
MirRelationExpr::Constant {
rows: Ok(vec![]),
typ,
},
)
}
/// Take ownership of `self`, leaving an empty `MirRelationExpr::Constant` with an **incorrect** type.
///
/// This should only be used if `self` is about to be dropped or otherwise overwritten.
pub fn take_dangerous(&mut self) -> MirRelationExpr {
let empty = MirRelationExpr::Constant {
rows: Ok(vec![]),
typ: RelationType::new(Vec::new()),
};
std::mem::replace(self, empty)
}
/// Replaces `self` with some logic applied to `self`.
pub fn replace_using<F>(&mut self, logic: F)
where
F: FnOnce(MirRelationExpr) -> MirRelationExpr,
{
let empty = MirRelationExpr::Constant {
rows: Ok(vec![]),
typ: RelationType::new(Vec::new()),
};
let expr = std::mem::replace(self, empty);
*self = logic(expr);
}
/// Store `self` in a `Let` and pass the corresponding `Get` to `body`
pub fn let_in<Body>(self, id_gen: &mut IdGen, body: Body) -> super::MirRelationExpr
where
Body: FnOnce(&mut IdGen, MirRelationExpr) -> super::MirRelationExpr,
{
if let MirRelationExpr::Get { .. } = self {
// already done
body(id_gen, self)
} else {
let id = LocalId::new(id_gen.allocate_id());
let get = MirRelationExpr::Get {
id: Id::Local(id),
typ: self.typ(),
};
let body = (body)(id_gen, get);
MirRelationExpr::Let {
id,
value: Box::new(self),
body: Box::new(body),
}
}
}
/// Return every row in `self` that does not have a matching row in the first columns of `keys_and_values`, using `default` to fill in the remaining columns
/// (If `default` is a row of nulls, this is the 'outer' part of LEFT OUTER JOIN)
pub fn anti_lookup(
self,
id_gen: &mut IdGen,
keys_and_values: MirRelationExpr,
default: Vec<(Datum, ColumnType)>,
) -> MirRelationExpr {
assert_eq!(keys_and_values.arity() - self.arity(), default.len());
self.let_in(id_gen, |_id_gen, get_keys| {
MirRelationExpr::join(
vec![
// all the missing keys (with count 1)
keys_and_values
.distinct_by((0..get_keys.arity()).collect())
.negate()
.union(get_keys.clone().distinct()),
// join with keys to get the correct counts
get_keys.clone(),
],
(0..get_keys.arity())
.map(|i| vec![(0, i), (1, i)])
.collect(),
)
// get rid of the extra copies of columns from keys
.project((0..get_keys.arity()).collect())
// This join is logically equivalent to
// `.map(<default_expr>)`, but using a join allows for
// potential predicate pushdown and elision in the
// optimizer.
.product(MirRelationExpr::constant(
vec![default.iter().map(|(datum, _)| *datum).collect()],
RelationType::new(default.iter().map(|(_, typ)| typ.clone()).collect()),
))
})
}
/// Return:
/// * every row in keys_and_values
/// * every row in `self` that does not have a matching row in the first columns of `keys_and_values`, using `default` to fill in the remaining columns
/// (This is LEFT OUTER JOIN if:
/// 1) `default` is a row of null
/// 2) matching rows in `keys_and_values` and `self` have the same multiplicity.)
pub fn lookup(
self,
id_gen: &mut IdGen,
keys_and_values: MirRelationExpr,
default: Vec<(Datum<'static>, ColumnType)>,
) -> MirRelationExpr {
keys_and_values.let_in(id_gen, |id_gen, get_keys_and_values| {
get_keys_and_values.clone().union(self.anti_lookup(
id_gen,
get_keys_and_values,
default,
))
})
}
/// Passes the collection through unchanged, but informs the optimizer that `keys` are primary keys.
pub fn declare_keys(self, keys: Vec<Vec<usize>>) -> Self {
Self::DeclareKeys {
input: Box::new(self),
keys,
}
}
/// Applies a fallible immutable `f` to each child of type `MirRelationExpr`.
pub fn try_visit_children<'a, F, E>(&'a self, f: F) -> Result<(), E>
where
F: FnMut(&'a MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_children(self, f)
}
/// Applies a fallible mutable `f` to each child of type `MirRelationExpr`.
pub fn try_visit_mut_children<'a, F, E>(&'a mut self, f: F) -> Result<(), E>
where
F: FnMut(&'a mut MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_mut_children(self, f)
}
/// Applies an infallible immutable `f` to each child of type `MirRelationExpr`.
pub fn visit_children<'a, F>(&'a self, f: F)
where
F: FnMut(&'a MirRelationExpr),
{
MirRelationExprVisitor::new().visit_children(self, f)
}
/// Applies an infallible mutable `f` to each child of type `MirRelationExpr`.
pub fn visit_mut_children<'a, F>(&'a mut self, f: F)
where
F: FnMut(&'a mut MirRelationExpr),
{
MirRelationExprVisitor::new().visit_mut_children(self, f)
}
/// Post-order immutable fallible `MirRelationExpr` visitor.
pub fn try_visit_post<'a, F, E>(&'a self, f: &mut F) -> Result<(), E>
where
F: FnMut(&'a MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_post(self, f)
}
/// Post-order mutable fallible `MirRelationExpr` visitor.
pub fn try_visit_mut_post<F, E>(&mut self, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_mut_post(self, f)
}
/// Post-order immutable infallible `MirRelationExpr` visitor.
pub fn visit_post<'a, F>(&'a self, f: &mut F)
where
F: FnMut(&'a MirRelationExpr),
{
MirRelationExprVisitor::new().visit_post(self, f)
}
/// Post-order mutable infallible `MirRelationExpr` visitor.
pub fn visit_mut_post<F>(&mut self, f: &mut F)
where
F: FnMut(&mut MirRelationExpr),
{
MirRelationExprVisitor::new().visit_mut_post(self, f)
}
/// Pre-order immutable fallible `MirRelationExpr` visitor.
pub fn try_visit_pre<F, E>(&self, f: &mut F) -> Result<(), E>
where
F: FnMut(&MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_pre(self, f)
}
/// Pre-order mutable fallible `MirRelationExpr` visitor.
pub fn try_visit_mut_pre<F, E>(&mut self, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_mut_pre(self, f)
}
/// Pre-order immutable infallible `MirRelationExpr` visitor.
pub fn visit_pre<F>(&self, f: &mut F)
where
F: FnMut(&MirRelationExpr),
{
MirRelationExprVisitor::new().visit_pre(self, f)
}
/// Pre-order mutable infallible `MirRelationExpr` visitor.
pub fn visit_mut_pre<F>(&mut self, f: &mut F)
where
F: FnMut(&mut MirRelationExpr),
{
MirRelationExprVisitor::new().visit_mut_pre(self, f)
}
/// A generalization of [`Self::visit_pre`] and [`Self::visit_post`].
///
/// The function `pre` runs on a `MirRelationExpr` before it runs on any of the
/// child `MirRelationExpr`s. The function `post` runs on child `MirRelationExpr`s
/// first before the parent.
///
/// Optionally, `pre` can return which child `MirRelationExpr`s, if any, should be
/// visited (default is to visit all children).
pub fn visit_pre_post<F1, F2>(&self, pre: &mut F1, post: &mut F2)
where
F1: FnMut(&MirRelationExpr) -> Option<Vec<&MirRelationExpr>>,
F2: FnMut(&MirRelationExpr),
{
MirRelationExprVisitor::new().visit_pre_post(self, pre, post)
}
/// Fallible visitor for the [`MirScalarExpr`]s directly owned by this relation expression.
///
/// The `f` visitor should not recursively descend into owned [`MirRelationExpr`]s.
pub fn try_visit_scalars_mut1<F, E>(&mut self, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirScalarExpr) -> Result<(), E>,
{
MirRelationExprVisitor::new().try_visit_scalar_children_mut(self, f)
}
/// Fallible mutable visitor for the [`MirScalarExpr`]s in the [`MirRelationExpr`] subtree rooted at `self`.
///
/// Note that this does not recurse into [`MirRelationExpr`] subtrees within [`MirScalarExpr`] nodes.
pub fn try_visit_scalars_mut<F, E>(&mut self, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirScalarExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
MirRelationExprVisitor::new().try_visit_scalars_mut(self, f)
}
/// Infallible mutable visitor for the [`MirScalarExpr`]s in the [`MirRelationExpr`] subtree rooted at at `self`.
///
/// Note that this does not recurse into [`MirRelationExpr`] subtrees within [`MirScalarExpr`] nodes.
pub fn visit_scalars_mut<F>(&mut self, f: &mut F)
where
F: FnMut(&mut MirScalarExpr),
{
MirRelationExprVisitor::new().visit_scalars_mut(self, f)
}
}
#[derive(Debug)]
struct MirRelationExprVisitor {
recursion_guard: RecursionGuard,
}
/// Contains visitor implementations.
///
/// [child, pre, post] x [fallible, infallible] x [immutable, mutable]
impl MirRelationExprVisitor {
/// Constructs a new MirRelationExprVisitor using a [`RecursionGuard`] with [`RECURSION_LIMIT`].
fn new() -> MirRelationExprVisitor {
MirRelationExprVisitor {
recursion_guard: RecursionGuard::with_limit(RECURSION_LIMIT),
}
}
/// Applies a fallible immutable `f` to each `expr` child of type `MirRelationExpr`.
fn try_visit_children<'a, F, E>(&self, expr: &'a MirRelationExpr, mut f: F) -> Result<(), E>
where
F: FnMut(&'a MirRelationExpr) -> Result<(), E>,
{
match expr {
MirRelationExpr::Constant { .. } | MirRelationExpr::Get { .. } => (),
MirRelationExpr::Let { value, body, .. } => {
f(value)?;
f(body)?;
}
MirRelationExpr::Project { input, .. } => {
f(input)?;
}
MirRelationExpr::Map { input, .. } => {
f(input)?;
}
MirRelationExpr::FlatMap { input, .. } => {
f(input)?;
}
MirRelationExpr::Filter { input, .. } => {
f(input)?;
}
MirRelationExpr::Join { inputs, .. } => {
for input in inputs {
f(input)?;
}
}
MirRelationExpr::Reduce { input, .. } => {
f(input)?;
}
MirRelationExpr::TopK { input, .. } => {
f(input)?;
}
MirRelationExpr::Negate { input } => f(input)?,
MirRelationExpr::Threshold { input } => f(input)?,
MirRelationExpr::Union { base, inputs } => {
f(base)?;
for input in inputs {
f(input)?;
}
}
MirRelationExpr::ArrangeBy { input, .. } => {
f(input)?;
}
MirRelationExpr::DeclareKeys { input, .. } => {
f(input)?;
}
}
Ok(())
}
/// Applies a fallible mutable `f` to each `expr` child of type `MirRelationExpr`.
fn try_visit_mut_children<'a, F, E>(
&self,
expr: &'a mut MirRelationExpr,
mut f: F,
) -> Result<(), E>
where
F: FnMut(&'a mut MirRelationExpr) -> Result<(), E>,
{
match expr {
MirRelationExpr::Constant { .. } | MirRelationExpr::Get { .. } => (),
MirRelationExpr::Let { value, body, .. } => {
f(value)?;
f(body)?;
}
MirRelationExpr::Project { input, .. } => {
f(input)?;
}
MirRelationExpr::Map { input, .. } => {
f(input)?;
}
MirRelationExpr::FlatMap { input, .. } => {
f(input)?;
}
MirRelationExpr::Filter { input, .. } => {
f(input)?;
}
MirRelationExpr::Join { inputs, .. } => {
for input in inputs {
f(input)?;
}
}
MirRelationExpr::Reduce { input, .. } => {
f(input)?;
}
MirRelationExpr::TopK { input, .. } => {
f(input)?;
}
MirRelationExpr::Negate { input } => f(input)?,
MirRelationExpr::Threshold { input } => f(input)?,
MirRelationExpr::Union { base, inputs } => {
f(base)?;
for input in inputs {
f(input)?;
}
}
MirRelationExpr::ArrangeBy { input, .. } => {
f(input)?;
}
MirRelationExpr::DeclareKeys { input, .. } => {
f(input)?;
}
}
Ok(())
}
/// Applies an infallible immutable `f` to each `expr` child of type `MirRelationExpr`.
fn visit_children<'a, F>(&self, expr: &'a MirRelationExpr, mut f: F)
where
F: FnMut(&'a MirRelationExpr),
{
match expr {
MirRelationExpr::Constant { .. } | MirRelationExpr::Get { .. } => (),
MirRelationExpr::Let { value, body, .. } => {
f(value);
f(body);
}
MirRelationExpr::Project { input, .. } => {
f(input);
}
MirRelationExpr::Map { input, .. } => {
f(input);
}
MirRelationExpr::FlatMap { input, .. } => {
f(input);
}
MirRelationExpr::Filter { input, .. } => {
f(input);
}
MirRelationExpr::Join { inputs, .. } => {
for input in inputs {
f(input);
}
}
MirRelationExpr::Reduce { input, .. } => {
f(input);
}
MirRelationExpr::TopK { input, .. } => {
f(input);
}
MirRelationExpr::Negate { input } => f(input),
MirRelationExpr::Threshold { input } => f(input),
MirRelationExpr::Union { base, inputs } => {
f(base);
for input in inputs {
f(input);
}
}
MirRelationExpr::ArrangeBy { input, .. } => {
f(input);
}
MirRelationExpr::DeclareKeys { input, .. } => {
f(input);
}
}
}
/// Applies an infallible mutable `f` to each `expr` child of type `MirRelationExpr`.
fn visit_mut_children<'a, F>(&self, expr: &'a mut MirRelationExpr, mut f: F)
where
F: FnMut(&'a mut MirRelationExpr),
{
match expr {
MirRelationExpr::Constant { .. } | MirRelationExpr::Get { .. } => (),
MirRelationExpr::Let { value, body, .. } => {
f(value);
f(body);
}
MirRelationExpr::Project { input, .. } => {
f(input);
}
MirRelationExpr::Map { input, .. } => {
f(input);
}
MirRelationExpr::FlatMap { input, .. } => {
f(input);
}
MirRelationExpr::Filter { input, .. } => {
f(input);
}
MirRelationExpr::Join { inputs, .. } => {
for input in inputs {
f(input);
}
}
MirRelationExpr::Reduce { input, .. } => {
f(input);
}
MirRelationExpr::TopK { input, .. } => {
f(input);
}
MirRelationExpr::Negate { input } => f(input),
MirRelationExpr::Threshold { input } => f(input),
MirRelationExpr::Union { base, inputs } => {
f(base);
for input in inputs {
f(input);
}
}
MirRelationExpr::ArrangeBy { input, .. } => {
f(input);
}
MirRelationExpr::DeclareKeys { input, .. } => {
f(input);
}
}
}
/// Post-order immutable fallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn try_visit_post<'a, F, E>(&self, expr: &'a MirRelationExpr, f: &mut F) -> Result<(), E>
where
F: FnMut(&'a MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
self.checked_recur(move |_| {
self.try_visit_children(expr, |e| self.try_visit_post(e, f))?;
f(expr)
})
}
/// Post-order mutable fallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn try_visit_mut_post<F, E>(&self, expr: &mut MirRelationExpr, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
self.checked_recur(move |_| {
self.try_visit_mut_children(expr, |e| self.try_visit_mut_post(e, f))?;
f(expr)
})
}
/// Post-order immutable infallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn visit_post<'a, F>(&self, expr: &'a MirRelationExpr, f: &mut F)
where
F: FnMut(&'a MirRelationExpr),
{
maybe_grow(|| {
self.visit_children(expr, |e| self.visit_post(e, f));
f(expr)
})
}
/// Post-order mutable infallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn visit_mut_post<F>(&self, expr: &mut MirRelationExpr, f: &mut F)
where
F: FnMut(&mut MirRelationExpr),
{
maybe_grow(|| {
self.visit_mut_children(expr, |e| self.visit_mut_post(e, f));
f(expr)
})
}
/// Pre-order immutable fallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn try_visit_pre<F, E>(&self, expr: &MirRelationExpr, f: &mut F) -> Result<(), E>
where
F: FnMut(&MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
self.checked_recur(move |_| {
f(expr)?;
self.try_visit_children(expr, |e| self.try_visit_pre(e, f))
})
}
/// Pre-order mutable fallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn try_visit_mut_pre<F, E>(&self, expr: &mut MirRelationExpr, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirRelationExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
self.checked_recur(move |_| {
f(expr)?;
self.try_visit_mut_children(expr, |e| self.try_visit_mut_pre(e, f))
})
}
/// Pre-order immutable infallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn visit_pre<F>(&self, expr: &MirRelationExpr, f: &mut F)
where
F: FnMut(&MirRelationExpr),
{
maybe_grow(|| {
f(expr);
self.visit_children(expr, |e| self.visit_pre(e, f))
})
}
/// Pre-order mutable infallible `MirRelationExpr` visitor for `expr`.
#[inline]
fn visit_mut_pre<F>(&self, expr: &mut MirRelationExpr, f: &mut F)
where
F: FnMut(&mut MirRelationExpr),
{
maybe_grow(|| {
f(expr);
self.visit_mut_children(expr, |e| self.visit_mut_pre(e, f))
})
}
/// A generalization of [`Self::visit_pre`] and [`Self::visit_post`].
///
/// The function `pre` runs on a `MirRelationExpr` before it runs on any of the
/// child `MirRelationExpr`s. The function `post` runs on child `MirRelationExpr`s
/// first before the parent.
///
/// Optionally, `pre` can return which child `MirRelationExpr`s, if any, should be
/// visited (default is to visit all children).
#[inline]
fn visit_pre_post<F1, F2>(&self, expr: &MirRelationExpr, pre: &mut F1, post: &mut F2)
where
F1: FnMut(&MirRelationExpr) -> Option<Vec<&MirRelationExpr>>,
F2: FnMut(&MirRelationExpr),
{
maybe_grow(|| {
if let Some(to_visit) = pre(expr) {
for e in to_visit {
self.visit_pre_post(e, pre, post);
}
} else {
self.visit_children(expr, |e| self.visit_pre_post(e, pre, post));
}
post(expr);
})
}
/// Fallible visitor for the [`MirScalarExpr`]s directly owned by this relation expression.
///
/// The `f` visitor should not recursively descend into owned [`MirRelationExpr`]s.
#[inline]
fn try_visit_scalar_children_mut<F, E>(
&self,
expr: &mut MirRelationExpr,
f: &mut F,
) -> Result<(), E>
where
F: FnMut(&mut MirScalarExpr) -> Result<(), E>,
{
// Match written out explicitly to reduce the possibility of adding a
// new field with a `MirScalarExpr` within and forgetting to account for it
// here.
match expr {
MirRelationExpr::Map { scalars, input: _ } => {
for s in scalars {
f(s)?;
}
Ok(())
}
MirRelationExpr::Filter {
predicates,
input: _,
} => {
for p in predicates {
f(p)?;
}
Ok(())
}
MirRelationExpr::FlatMap {
exprs,
input: _,
func: _,
} => {
for expr in exprs {
f(expr)?;
}
Ok(())
}
MirRelationExpr::Join {
equivalences,
inputs: _,
implementation: _,
} => {
for equivalence in equivalences {
for expr in equivalence {
f(expr)?;
}
}
Ok(())
}
MirRelationExpr::ArrangeBy { input: _, keys } => {
for key in keys {
for s in key {
f(s)?;
}
}
Ok(())
}
MirRelationExpr::Reduce {
group_key,
aggregates,
..
} => {
for s in group_key {
f(s)?;
}
for agg in aggregates {
f(&mut agg.expr)?;
}
Ok(())
}
MirRelationExpr::Constant { rows: _, typ: _ }
| MirRelationExpr::Get { id: _, typ: _ }
| MirRelationExpr::Let {
id: _,
value: _,
body: _,
}
| MirRelationExpr::Project {
input: _,
outputs: _,
}
| MirRelationExpr::TopK {
input: _,
group_key: _,
order_key: _,
limit: _,
offset: _,
monotonic: _,
}
| MirRelationExpr::Negate { input: _ }
| MirRelationExpr::Threshold { input: _ }
| MirRelationExpr::DeclareKeys { input: _, keys: _ }
| MirRelationExpr::Union { base: _, inputs: _ } => Ok(()),
}
}
/// Fallible mutable visitor for all [`MirScalarExpr`]s in the [`MirRelationExpr`] subtree rooted at `expr`.
///
/// Note that this does not recurse into [`MirRelationExpr`] subtrees wrapped in [`MirScalarExpr`] nodes.
#[inline]
fn try_visit_scalars_mut<F, E>(&self, expr: &mut MirRelationExpr, f: &mut F) -> Result<(), E>
where
F: FnMut(&mut MirScalarExpr) -> Result<(), E>,
E: From<RecursionLimitError>,
{
self.try_visit_mut_post(expr, &mut |e| self.try_visit_scalar_children_mut(e, f))
}
/// Infallible mutable visitor for the [`MirScalarExpr`]s in the [`MirRelationExpr`] subtree rooted at `expr`.
///
/// Note that this does not recurse into [`MirRelationExpr`] subtrees within [`MirScalarExpr`] nodes.
#[inline]
fn visit_scalars_mut<F>(&self, expr: &mut MirRelationExpr, f: &mut F)
where
F: FnMut(&mut MirScalarExpr),
{
self.try_visit_scalars_mut(expr, &mut |s| {
f(s);
Ok::<_, RecursionLimitError>(())
})
.expect("Unexpected error in `visit_scalars_mut` call")
}
}
/// Add checked recursion support for [`MirRelationExprVisitor`].
impl CheckedRecursion for MirRelationExprVisitor {
fn recursion_guard(&self) -> &RecursionGuard {
&self.recursion_guard
}
}
/// Specification for an ordering by a column.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Hash, MzReflect)]
pub struct ColumnOrder {
/// The column index.
pub column: usize,
/// Whether to sort in descending order.
#[serde(default)]
pub desc: bool,
}
impl fmt::Display for ColumnOrder {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(
f,
"#{} {}",
self.column,
if self.desc { "desc" } else { "asc" }
)
}
}
/// Describes an aggregation expression.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize, Hash, MzReflect)]
pub struct AggregateExpr {
/// Names the aggregation function.
pub func: AggregateFunc,
/// An expression which extracts from each row the input to `func`.
pub expr: MirScalarExpr,
/// Should the aggregation be applied only to distinct results in each group.
#[serde(default)]
pub distinct: bool,
}
impl AggregateExpr {
/// Computes the type of this `AggregateExpr`.
pub fn typ(&self, relation_type: &RelationType) -> ColumnType {
self.func.output_type(self.expr.typ(relation_type))
}
/// Returns whether the expression has a constant result.
pub fn is_constant(&self) -> bool {
match self.func {
AggregateFunc::MaxInt16
| AggregateFunc::MaxInt32
| AggregateFunc::MaxInt64
| AggregateFunc::MaxFloat32
| AggregateFunc::MaxFloat64
| AggregateFunc::MaxBool
| AggregateFunc::MaxString
| AggregateFunc::MaxDate
| AggregateFunc::MaxTimestamp
| AggregateFunc::MaxTimestampTz
| AggregateFunc::MinInt16
| AggregateFunc::MinInt32
| AggregateFunc::MinInt64
| AggregateFunc::MinFloat32
| AggregateFunc::MinFloat64
| AggregateFunc::MinBool
| AggregateFunc::MinString
| AggregateFunc::MinDate
| AggregateFunc::MinTimestamp
| AggregateFunc::MinTimestampTz
| AggregateFunc::Any
| AggregateFunc::All
| AggregateFunc::Dummy => self.expr.is_literal(),
AggregateFunc::Count => self.expr.is_literal_null(),
_ => self.expr.is_literal_err(),
}
}
/// Extracts unique input from aggregate type
pub fn on_unique(&self, input_type: &RelationType) -> MirScalarExpr {
match self.func {
// Count is one if non-null, and zero if null.
AggregateFunc::Count => self
.expr
.clone()
.call_unary(UnaryFunc::IsNull(crate::func::IsNull))
.if_then_else(
MirScalarExpr::literal_ok(Datum::Int64(0), ScalarType::Int64),
MirScalarExpr::literal_ok(Datum::Int64(1), ScalarType::Int64),
),
// SumInt16 takes Int16s as input, but outputs Int64s.
AggregateFunc::SumInt16 => self
.expr
.clone()
.call_unary(UnaryFunc::CastInt16ToInt64(scalar_func::CastInt16ToInt64)),
// SumInt32 takes Int32s as input, but outputs Int64s.
AggregateFunc::SumInt32 => self
.expr
.clone()
.call_unary(UnaryFunc::CastInt32ToInt64(scalar_func::CastInt32ToInt64)),
// SumInt64 takes Int64s as input, but outputs numerics.
AggregateFunc::SumInt64 => self.expr.clone().call_unary(UnaryFunc::CastInt64ToNumeric(
scalar_func::CastInt64ToNumeric(Some(0)),
)),
// JsonbAgg takes _anything_ as input, but must output a Jsonb array.
AggregateFunc::JsonbAgg { .. } => MirScalarExpr::CallVariadic {
func: VariadicFunc::JsonbBuildArray,
exprs: vec![self.expr.clone().call_unary(UnaryFunc::RecordGet(0))],
},
// JsonbAgg takes _anything_ as input, but must output a Jsonb object.
AggregateFunc::JsonbObjectAgg { .. } => {
let record = self.expr.clone().call_unary(UnaryFunc::RecordGet(0));
MirScalarExpr::CallVariadic {
func: VariadicFunc::JsonbBuildObject,
exprs: (0..2)
.map(|i| record.clone().call_unary(UnaryFunc::RecordGet(i)))
.collect(),
}
}
// StringAgg takes nested records of strings and outputs a string
AggregateFunc::StringAgg { .. } => self
.expr
.clone()
.call_unary(UnaryFunc::RecordGet(0))
.call_unary(UnaryFunc::RecordGet(0)),
// ListConcat and ArrayConcat take a single level of records and output a list containing exactly 1 element
AggregateFunc::ListConcat { .. } | AggregateFunc::ArrayConcat { .. } => {
self.expr.clone().call_unary(UnaryFunc::RecordGet(0))
}
// RowNumber takes a list of records and outputs a list containing exactly 1 element
AggregateFunc::RowNumber { .. } => {
let record = self
.expr
.clone()
// extract the list within the record
.call_unary(UnaryFunc::RecordGet(0))
// extract the expression within the list
.call_binary(
MirScalarExpr::literal_ok(Datum::Int64(1), ScalarType::Int64),
BinaryFunc::ListIndex,
);
MirScalarExpr::CallVariadic {
func: VariadicFunc::ListCreate {
elem_type: self
.typ(input_type)
.scalar_type
.unwrap_list_element_type()
.clone(),
},
exprs: vec![MirScalarExpr::CallVariadic {
func: VariadicFunc::RecordCreate {
field_names: vec![
ColumnName::from("?row_number?"),
ColumnName::from("?record?"),
],
},
exprs: vec![
MirScalarExpr::literal_ok(Datum::Int64(1), ScalarType::Int64),
record,
],
}],
}
}
// All other variants should return the argument to the aggregation.
AggregateFunc::MaxNumeric
| AggregateFunc::MaxInt16
| AggregateFunc::MaxInt32
| AggregateFunc::MaxInt64
| AggregateFunc::MaxFloat32
| AggregateFunc::MaxFloat64
| AggregateFunc::MaxBool
| AggregateFunc::MaxString
| AggregateFunc::MaxDate
| AggregateFunc::MaxTimestamp
| AggregateFunc::MaxTimestampTz
| AggregateFunc::MinNumeric
| AggregateFunc::MinInt16
| AggregateFunc::MinInt32
| AggregateFunc::MinInt64
| AggregateFunc::MinFloat32
| AggregateFunc::MinFloat64
| AggregateFunc::MinBool
| AggregateFunc::MinString
| AggregateFunc::MinDate
| AggregateFunc::MinTimestamp
| AggregateFunc::MinTimestampTz
| AggregateFunc::SumFloat32
| AggregateFunc::SumFloat64
| AggregateFunc::SumNumeric
| AggregateFunc::Any
| AggregateFunc::All
| AggregateFunc::Dummy => self.expr.clone(),
}
}
}
impl fmt::Display for AggregateExpr {
fn fmt(&self, f: &mut fmt::Formatter) -> Result<(), fmt::Error> {
write!(
f,
"{}({}{})",
self.func,
if self.distinct { "distinct " } else { "" },
self.expr
)
}
}
/// Describe a join implementation in dataflow.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize, Hash, MzReflect)]
pub enum JoinImplementation {
/// Perform a sequence of binary differential dataflow joins.
///
/// The first argument indicates 1) the index of the starting collection
/// and 2) if it should be arranged, the keys to arrange it by.
/// The sequence that follows lists other relation indexes, and the key for
/// the arrangement we should use when joining it in.
///
/// Each collection index should occur exactly once, either in the first
/// position or somewhere in the list.
Differential(
(usize, Option<Vec<MirScalarExpr>>),
Vec<(usize, Vec<MirScalarExpr>)>,
),
/// Perform independent delta query dataflows for each input.
///
/// The argument is a sequence of plans, for the input collections in order.
/// Each plan starts from the corresponding index, and then in sequence joins
/// against collections identified by index and with the specified arrangement key.
DeltaQuery(Vec<Vec<(usize, Vec<MirScalarExpr>)>>),
/// No implementation yet selected.
Unimplemented,
}
impl Default for JoinImplementation {
fn default() -> Self {
JoinImplementation::Unimplemented
}
}
/// Instructions for finishing the result of a query.
///
/// The primary reason for the existence of this structure and attendant code
/// is that SQL's ORDER BY requires sorting rows (as already implied by the
/// keywords), whereas much of the rest of SQL is defined in terms of unordered
/// multisets. But as it turns out, the same idea can be used to optimize
/// trivial peeks.
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub struct RowSetFinishing {
/// Order rows by the given columns.
pub order_by: Vec<ColumnOrder>,
/// Include only as many rows (after offset).
pub limit: Option<usize>,
/// Omit as many rows.
pub offset: usize,
/// Include only given columns.
pub project: Vec<usize>,
}
impl RowSetFinishing {
/// True if the finishing does nothing to any result set.
pub fn is_trivial(&self, arity: usize) -> bool {
self.limit.is_none()
&& self.order_by.is_empty()
&& self.offset == 0
&& self.project.iter().copied().eq(0..arity)
}
/// Applies finishing actions to a row set.
pub fn finish(&self, rows: &mut Vec<Row>) {
let mut left_datum_vec = repr::DatumVec::new();
let mut right_datum_vec = repr::DatumVec::new();
let mut sort_by = |left: &Row, right: &Row| {
let left_datums = left_datum_vec.borrow_with(left);
let right_datums = right_datum_vec.borrow_with(right);
compare_columns(&self.order_by, &left_datums, &right_datums, || {
left.cmp(&right)
})
};
let offset = self.offset;
if offset > rows.len() {
*rows = Vec::new();
} else {
if let Some(limit) = self.limit {
let offset_plus_limit = offset + limit;
if rows.len() > offset_plus_limit {
pdqselect::select_by(rows, offset_plus_limit, &mut sort_by);
rows.truncate(offset_plus_limit);
}
}
if offset > 0 {
pdqselect::select_by(rows, offset, &mut sort_by);
rows.drain(..offset);
}
rows.sort_by(&mut sort_by);
let mut row_packer = Row::default();
let mut datum_vec = repr::DatumVec::new();
for row in rows.iter_mut() {
*row = {
let datums = datum_vec.borrow_with(&row);
row_packer.extend(self.project.iter().map(|i| &datums[*i]));
row_packer.finish_and_reuse()
};
}
}
}
}
/// Compare `left` and `right` using `order`. If that doesn't produce a strict ordering, call `tiebreaker`.
pub fn compare_columns<F>(
order: &[ColumnOrder],
left: &[Datum],
right: &[Datum],
tiebreaker: F,
) -> Ordering
where
F: Fn() -> Ordering,
{
for order in order {
let (lval, rval) = (&left[order.column], &right[order.column]);
let cmp = if order.desc {
rval.cmp(&lval)
} else {
lval.cmp(&rval)
};
if cmp != Ordering::Equal {
return cmp;
}
}
tiebreaker()
}