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
// 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.
//! The types for the dataflow crate.
//!
//! These are extracted into their own crate so that crates that only depend
//! on the interface of the dataflow crate, and not its implementation, can
//! avoid the dependency, as the dataflow crate is very slow to compile.
use std::collections::{BTreeMap, HashSet};
use serde::{Deserialize, Serialize};
use timely::progress::frontier::Antichain;
use expr::{GlobalId, MirRelationExpr, MirScalarExpr, OptimizedMirRelationExpr};
use repr::{Diff, RelationType, Row, Timestamp};
use crate::sources::persistence::SourcePersistDesc;
use crate::types::sources::SourceDesc;
/// The response from a `Peek`.
///
/// Note that each `Peek` expects to generate exactly one `PeekResponse`, i.e.
/// we expect a 1:1 contract between `Peek` and `PeekResponse`.
#[derive(Clone, Debug, Serialize, Deserialize, PartialEq)]
pub enum PeekResponse {
Rows(Vec<Row>),
Error(String),
Canceled,
}
impl PeekResponse {
pub fn unwrap_rows(self) -> Vec<Row> {
match self {
PeekResponse::Rows(rows) => rows,
PeekResponse::Error(_) | PeekResponse::Canceled => {
panic!("PeekResponse::unwrap_rows called on {:?}", self)
}
}
}
}
/// Various responses that can be communicated about the progress of a TAIL command.
#[derive(Debug, Clone, Eq, PartialEq, Serialize, Deserialize)]
pub enum TailResponse {
/// Progress information. Subsequent messages from this worker will only contain timestamps
/// greater or equal to an element of this frontier.
///
/// An empty antichain indicates the end.
Progress(Antichain<Timestamp>),
/// Rows that should be returned in order to the client.
Rows(Vec<(Row, Timestamp, Diff)>),
/// The TAIL dataflow was dropped before completing. Indicates the end.
Dropped,
}
#[derive(Debug, Clone, Eq, PartialEq, Serialize, Deserialize)]
/// A batch of updates to be fed to a local input
pub struct Update {
pub row: Row,
pub timestamp: u64,
pub diff: Diff,
}
/// A commonly used name for dataflows contain MIR expressions.
pub type DataflowDesc = DataflowDescription<OptimizedMirRelationExpr>;
/// An association of a global identifier to an expression.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct BuildDesc<View> {
pub id: GlobalId,
pub view: View,
}
/// A description of an instantation of a source.
///
/// This includes a description of the source, but additionally any
/// context-dependent options like the ability to apply filtering and
/// projection to the records as they emerge.
#[derive(Debug, Clone, Eq, PartialEq, Serialize, Deserialize)]
pub struct SourceInstanceDesc {
/// A description of the source to construct.
pub description: crate::types::sources::SourceDesc,
/// Optional linear operators that can be applied record-by-record.
pub operators: Option<crate::types::LinearOperator>,
/// A description of how to persist the source.
pub persist: Option<sources::persistence::SourcePersistDesc>,
}
/// A description of a dataflow to construct and results to surface.
#[derive(Clone, Debug, Default, Serialize, Deserialize)]
pub struct DataflowDescription<View> {
/// Sources instantiations made available to the dataflow.
pub source_imports: BTreeMap<GlobalId, SourceInstanceDesc>,
/// Indexes made available to the dataflow.
pub index_imports: BTreeMap<GlobalId, (IndexDesc, RelationType)>,
/// Views and indexes to be built and stored in the local context.
/// Objects must be built in the specific order, as there may be
/// dependencies of later objects on prior identifiers.
pub objects_to_build: Vec<BuildDesc<View>>,
/// Indexes to be made available to be shared with other dataflows
/// (id of new index, description of index, relationtype of base source/view)
pub index_exports: Vec<(GlobalId, IndexDesc, RelationType)>,
/// sinks to be created
/// (id of new sink, description of sink)
pub sink_exports: Vec<(GlobalId, crate::types::sinks::SinkDesc)>,
/// Maps views to views + indexes needed to generate that view
pub dependent_objects: BTreeMap<GlobalId, Vec<GlobalId>>,
/// An optional frontier to which inputs should be advanced.
///
/// If this is set, it should override the default setting determined by
/// the upper bound of `since` frontiers contributing to the dataflow.
/// It is an error for this to be set to a frontier not beyond that default.
pub as_of: Option<Antichain<Timestamp>>,
/// Human readable name
pub debug_name: String,
}
impl DataflowDescription<OptimizedMirRelationExpr> {
/// Creates a new dataflow description with a human-readable name.
pub fn new(name: String) -> Self {
Self {
source_imports: Default::default(),
index_imports: Default::default(),
objects_to_build: Vec::new(),
index_exports: Default::default(),
sink_exports: Default::default(),
dependent_objects: Default::default(),
as_of: Default::default(),
debug_name: name,
}
}
/// Imports a previously exported index.
///
/// This method makes available an index previously exported as `id`, identified
/// to the query by `description` (which names the view the index arranges, and
/// the keys by which it is arranged).
///
/// The `requesting_view` argument is currently necessary to correctly track the
/// dependencies of views on indexes.
pub fn import_index(
&mut self,
id: GlobalId,
description: IndexDesc,
typ: RelationType,
requesting_view: GlobalId,
) {
self.index_imports.insert(id, (description, typ));
self.record_depends_on(requesting_view, id);
}
/// Imports a source and makes it available as `id`.
pub fn import_source(
&mut self,
id: GlobalId,
description: SourceDesc,
persist: Option<SourcePersistDesc>,
) {
// Import the source with no linear operators applied to it.
// They may be populated by whole-dataflow optimization.
self.source_imports.insert(
id,
SourceInstanceDesc {
description,
operators: None,
persist,
},
);
}
/// Binds to `id` the relation expression `view`.
pub fn insert_view(&mut self, id: GlobalId, view: OptimizedMirRelationExpr) {
for get_id in view.global_uses() {
self.record_depends_on(id, get_id);
}
self.objects_to_build.push(BuildDesc { id, view });
}
/// Exports as `id` an index on `on_id`.
///
/// Future uses of `import_index` in other dataflow descriptions may use `id`,
/// as long as this dataflow has not been terminated in the meantime.
pub fn export_index(&mut self, id: GlobalId, description: IndexDesc, on_type: RelationType) {
// We first create a "view" named `id` that ensures that the
// data are correctly arranged and available for export.
self.insert_view(
id,
OptimizedMirRelationExpr::declare_optimized(MirRelationExpr::ArrangeBy {
input: Box::new(MirRelationExpr::global_get(
description.on_id,
on_type.clone(),
)),
keys: vec![description.key.clone()],
}),
);
self.index_exports.push((id, description, on_type));
}
/// Exports as `id` a sink described by `description`.
pub fn export_sink(&mut self, id: GlobalId, description: crate::types::sinks::SinkDesc) {
self.sink_exports.push((id, description));
}
/// Records a dependency of `view_id` on `depended_upon`.
// TODO(#7267): This information should ideally be automatically extracted
// from the imported sources and indexes, rather than relying on the caller
// to correctly specify them.
fn record_depends_on(&mut self, view_id: GlobalId, depended_upon: GlobalId) {
self.dependent_objects
.entry(view_id)
.or_insert_with(Vec::new)
.push(depended_upon);
}
/// Returns true iff `id` is already imported.
pub fn is_imported(&self, id: &GlobalId) -> bool {
self.objects_to_build.iter().any(|bd| &bd.id == id)
|| self.source_imports.iter().any(|(i, _)| i == id)
}
/// Assigns the `as_of` frontier to the supplied argument.
///
/// This method allows the dataflow to indicate a frontier up through
/// which all times should be advanced. This can be done for at least
/// two reasons: 1. correctness and 2. performance.
///
/// Correctness may require an `as_of` to ensure that historical detail
/// is consolidated at representative times that do not present specific
/// detail that is not specifically correct. For example, updates may be
/// compacted to times that are no longer the source times, but instead
/// some byproduct of when compaction was executed; we should not present
/// those specific times as meaningfully different from other equivalent
/// times.
///
/// Performance may benefit from an aggressive `as_of` as it reduces the
/// number of distinct moments at which collections vary. Differential
/// dataflow will refresh its outputs at each time its inputs change and
/// to moderate that we can minimize the volume of distinct input times
/// as much as possible.
///
/// Generally, one should consider setting `as_of` at least to the `since`
/// frontiers of contributing data sources and as aggressively as the
/// computation permits.
pub fn set_as_of(&mut self, as_of: Antichain<Timestamp>) {
self.as_of = Some(as_of);
}
/// The number of columns associated with an identifier in the dataflow.
pub fn arity_of(&self, id: &GlobalId) -> usize {
for (source_id, source) in self.source_imports.iter() {
if source_id == id {
return source.description.desc.arity();
}
}
for (_index_id, (desc, typ)) in self.index_imports.iter() {
if &desc.on_id == id {
return typ.arity();
}
}
for desc in self.objects_to_build.iter() {
if &desc.id == id {
return desc.view.arity();
}
}
panic!("GlobalId {} not found in DataflowDesc", id);
}
}
impl<View> DataflowDescription<View> {
/// Collects all indexes required to construct all exports.
pub fn get_all_imports(&self) -> HashSet<GlobalId> {
let mut result = HashSet::new();
for (_, desc, _) in &self.index_exports {
result.extend(self.get_imports(&desc.on_id))
}
for (_, sink) in &self.sink_exports {
result.extend(self.get_imports(&sink.from))
}
result
}
/// Collects all transitively dependent identifiers that do not have their own dependencies.
pub fn get_imports(&self, id: &GlobalId) -> HashSet<GlobalId> {
let mut result = HashSet::new();
let mut worklist = vec![id];
while let Some(id) = worklist.pop() {
result.insert(*id);
if let Some(dependents) = self.dependent_objects.get(id) {
for id in dependents.iter() {
if !result.contains(id) {
worklist.push(id);
}
}
}
}
result.retain(|id| self.dependent_objects.get(id).is_none());
result
}
}
/// Types and traits related to the introduction of changing collections into `dataflow`.
pub mod sources {
use std::collections::{BTreeMap, HashMap};
use std::ops::Add;
use std::path::PathBuf;
use std::time::Duration;
use anyhow::{anyhow, bail};
use chrono::NaiveDateTime;
use globset::Glob;
use http::Uri;
use serde::{Deserialize, Serialize};
use uuid::Uuid;
use kafka_util::KafkaAddrs;
use repr::{ColumnType, RelationDesc, RelationType, ScalarType};
// Types and traits related to the *decoding* of data for sources.
pub mod encoding {
use anyhow::Context;
use serde::{Deserialize, Serialize};
use interchange::{avro, protobuf};
use repr::{ColumnType, RelationDesc, ScalarType};
/// A description of how to interpret data from various sources
///
/// Almost all sources only present values as part of their records, but Kafka allows a key to be
/// associated with each record, which has a possibly independent encoding.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum SourceDataEncoding {
Single(DataEncoding),
KeyValue {
key: DataEncoding,
value: DataEncoding,
},
}
/// A description of how each row should be decoded, from a string of bytes to a sequence of
/// Differential updates.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum DataEncoding {
Avro(AvroEncoding),
AvroOcf(AvroOcfEncoding),
Protobuf(ProtobufEncoding),
Csv(CsvEncoding),
Regex(RegexEncoding),
Postgres,
Bytes,
Text,
}
impl SourceDataEncoding {
pub fn key_ref(&self) -> Option<&DataEncoding> {
match self {
SourceDataEncoding::Single(_) => None,
SourceDataEncoding::KeyValue { key, .. } => Some(key),
}
}
/// Return either the Single encoding if this was a `SourceDataEncoding::Single`, else return the value encoding
pub fn value(self) -> DataEncoding {
match self {
SourceDataEncoding::Single(encoding) => encoding,
SourceDataEncoding::KeyValue { value, .. } => value,
}
}
pub fn value_ref(&self) -> &DataEncoding {
match self {
SourceDataEncoding::Single(encoding) => encoding,
SourceDataEncoding::KeyValue { value, .. } => value,
}
}
pub fn desc(&self) -> Result<(Option<RelationDesc>, RelationDesc), anyhow::Error> {
Ok(match self {
SourceDataEncoding::Single(value) => (None, value.desc()?),
SourceDataEncoding::KeyValue { key, value } => {
(Some(key.desc()?), value.desc()?)
}
})
}
}
pub fn included_column_desc(included_columns: Vec<(&str, ColumnType)>) -> RelationDesc {
let mut desc = RelationDesc::empty();
for (name, ty) in included_columns {
desc = desc.with_column(name, ty);
}
desc
}
impl DataEncoding {
/// Computes the [`RelationDesc`] for the relation specified by this
/// data encoding and envelope.
///
/// If a key desc is provided it will be prepended to the returned desc
fn desc(&self) -> Result<RelationDesc, anyhow::Error> {
// Add columns for the data, based on the encoding format.
Ok(match self {
DataEncoding::Bytes => {
RelationDesc::empty().with_column("data", ScalarType::Bytes.nullable(false))
}
DataEncoding::AvroOcf(AvroOcfEncoding {
reader_schema: schema,
..
})
| DataEncoding::Avro(AvroEncoding { schema, .. }) => {
let parsed_schema =
avro::parse_schema(schema).context("validating avro schema")?;
avro::schema_to_relationdesc(parsed_schema)
.context("validating avro schema")?
}
DataEncoding::Protobuf(ProtobufEncoding {
descriptors,
message_name,
confluent_wire_format: _,
}) => protobuf::DecodedDescriptors::from_bytes(
descriptors,
message_name.to_owned(),
)?
.columns()
.iter()
.fold(RelationDesc::empty(), |desc, (name, ty)| {
desc.with_column(name, ty.clone())
}),
DataEncoding::Regex(RegexEncoding { regex }) => regex
.capture_names()
.enumerate()
// The first capture is the entire matched string. This will
// often not be useful, so skip it. If people want it they can
// just surround their entire regex in an explicit capture
// group.
.skip(1)
.fold(RelationDesc::empty(), |desc, (i, name)| {
let name = match name {
None => format!("column{}", i),
Some(name) => name.to_owned(),
};
let ty = ScalarType::String.nullable(true);
desc.with_column(name, ty)
}),
DataEncoding::Csv(CsvEncoding { columns, .. }) => match columns {
ColumnSpec::Count(n) => {
(1..=*n).into_iter().fold(RelationDesc::empty(), |desc, i| {
desc.with_column(
format!("column{}", i),
ScalarType::String.nullable(false),
)
})
}
ColumnSpec::Header { names } => names
.iter()
.map(|s| &**s)
.fold(RelationDesc::empty(), |desc, name| {
desc.with_column(name, ScalarType::String.nullable(false))
}),
},
DataEncoding::Text => RelationDesc::empty()
.with_column("text", ScalarType::String.nullable(false)),
DataEncoding::Postgres => RelationDesc::empty()
.with_column("oid", ScalarType::Int32.nullable(false))
.with_column(
"row_data",
ScalarType::List {
element_type: Box::new(ScalarType::String),
custom_oid: None,
}
.nullable(false),
),
})
}
pub fn op_name(&self) -> &'static str {
match self {
DataEncoding::Bytes => "Bytes",
DataEncoding::AvroOcf { .. } => "AvroOcf",
DataEncoding::Avro(_) => "Avro",
DataEncoding::Protobuf(_) => "Protobuf",
DataEncoding::Regex { .. } => "Regex",
DataEncoding::Csv(_) => "Csv",
DataEncoding::Text => "Text",
DataEncoding::Postgres => "Postgres",
}
}
}
/// Encoding in Avro format.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct AvroEncoding {
pub schema: String,
pub schema_registry_config: Option<ccsr::ClientConfig>,
pub confluent_wire_format: bool,
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct AvroOcfEncoding {
pub reader_schema: String,
}
/// Encoding in Protobuf format.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct ProtobufEncoding {
pub descriptors: Vec<u8>,
pub message_name: String,
pub confluent_wire_format: bool,
}
/// Arguments necessary to define how to decode from CSV format
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct CsvEncoding {
pub columns: ColumnSpec,
pub delimiter: u8,
}
/// Determines the RelationDesc and decoding of CSV objects
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum ColumnSpec {
/// The first row is not a header row, and all columns get default names like `columnN`.
Count(usize),
/// The first row is a header row and therefore does become data
///
/// Each of the values in `names` becomes the default name of a column in the dataflow.
Header { names: Vec<String> },
}
impl ColumnSpec {
/// The number of columns described by the column spec.
pub fn arity(&self) -> usize {
match self {
ColumnSpec::Count(n) => *n,
ColumnSpec::Header { names } => names.len(),
}
}
pub fn into_header_names(self) -> Option<Vec<String>> {
match self {
ColumnSpec::Count(_) => None,
ColumnSpec::Header { names } => Some(names),
}
}
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct RegexEncoding {
pub regex: repr::adt::regex::Regex,
}
}
pub mod persistence {
use serde::{Deserialize, Serialize};
use expr::PartitionId;
/// The details needed to make a source that uses an external [`super::SourceConnector`] persistent.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct SourcePersistDesc {
/// The _current_ upper seal timestamp of all involved streams.
///
/// NOTE: This timestamp is determined when the coordinator starts up or when the source is
/// initially created. When a source is actively writing to this stream, the seal timestamp
/// will progress beyond this timestamp.
///
/// This is okay for now because we only want to allow one source instantiation for persistent
/// sources, meaning the flow is usually this:
///
/// 1. coordinator determines seal timestamp
/// 2. seal timestamps for a source are sent to dataflow when rendering a source
/// 3. coordinator (or anyone) never looks at this timestamp again.
///
/// And when we restart, we start from step 1., at which time we are guaranteed not to have a
/// source running already.
pub upper_seal_ts: u64,
/// The _current_ compaction frontier (aka _since_) of all involved streams.
///
/// NOTE: This timestamp is determined when the coordinator starts up or when the source is
/// initially created. When a source is actively writing to this stream and allowing
/// compaction, this will progress beyond this timestamp.
///
/// This is okay for now because we only want to allow one source instantiation for persistent
/// sources, meaning the flow is usually this:
///
/// 1. coordinator determines since timestamp
/// 2. timestamps for a source are sent to dataflow when rendering a source
/// 3. coordinator (or anyone) never looks at this timestamp again.
///
/// And when we restart, we start from step 1., at which time we are guaranteed not to have a
/// source running already.
pub since_ts: u64,
/// Name of the primary persisted stream of this source. This is what a consumer of the
/// persisted data would be interested in while the secondary stream(s) of the source are an
/// internal implementation detail.
pub primary_stream: String,
/// Persisted stream of timestamp bindings.
pub timestamp_bindings_stream: String,
/// Any additional details that we need to make the envelope logic stateful.
pub envelope_desc: EnvelopePersistDesc,
}
/// The persistence details we need for persisting a source envelopes data structures.
///
/// This is a 1:1 mapping from envelope to `EnvelopePersistDesc`, as opposed to having a `None`
/// that covers all envelopes that don't need additional data. Mostly, to just be explicit, but
/// also because there is already a `NONE` envelope.
///
/// Some envelopes will require additional streams, which should be listed in the variant.
#[derive(Debug, Clone, Serialize, Deserialize, Eq, PartialEq)]
pub enum EnvelopePersistDesc {
Upsert,
None,
}
/// Structure wrapping a timestamp update from a source
/// If RT, contains a partition count
/// which informs workers that messages with Offset on PartititionId will be timestamped
/// with Timestamp.
#[derive(Clone, Debug, Serialize, Deserialize)]
pub enum TimestampSourceUpdate {
/// Update for an RT source: contains a new partition to add to this source.
RealTime(PartitionId),
}
}
/// Universal language for describing message positions in Materialize, in a source independent
/// way. Invidual sources like Kafka or File sources should explicitly implement their own offset
/// type that converts to/From MzOffsets. A 0-MzOffset denotes an empty stream.
#[derive(Copy, Clone, Debug, PartialEq, PartialOrd, Eq, Hash, Serialize, Deserialize)]
pub struct MzOffset {
pub offset: i64,
}
impl std::fmt::Display for MzOffset {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
write!(f, "{}", self.offset)
}
}
impl Add<i64> for MzOffset {
type Output = MzOffset;
fn add(self, x: i64) -> MzOffset {
MzOffset {
offset: self.offset + x,
}
}
}
#[derive(Clone, Copy, Eq, PartialEq)]
pub struct KafkaOffset {
pub offset: i64,
}
/// Convert from KafkaOffset to MzOffset (1-indexed)
impl From<KafkaOffset> for MzOffset {
fn from(kafka_offset: KafkaOffset) -> Self {
MzOffset {
offset: kafka_offset.offset + 1,
}
}
}
/// Convert from MzOffset to Kafka::Offset as long as
/// the offset is not negative
impl Into<KafkaOffset> for MzOffset {
fn into(self) -> KafkaOffset {
KafkaOffset {
offset: self.offset - 1,
}
}
}
/// Which piece of metadata a column corresponds to
#[derive(Copy, Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum IncludedColumnSource {
/// The materialize-specific notion of "position"
///
/// This is legacy, and should be removed when default metadata is no longer included
DefaultPosition,
Partition,
Offset,
Timestamp,
Topic,
}
/// Whether and how to include the decoded key of a stream in dataflows
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum KeyEnvelope {
/// Never include the key in the output row
None,
/// For composite key encodings, pull the fields from the encoding into columns.
Flattened,
/// Upsert is identical to Flattened but differs for non-avro sources, for which key names are overwritten.
LegacyUpsert,
/// Always use the given name for the key.
///
/// * For a single-field key, this means that the column will get the given name.
/// * For a multi-column key, the columns will get packed into a [`ScalarType::Record`], and
/// that Record will get the given name.
Named(String),
}
/// A column that was created via an `INCLUDE` expression
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct IncludedColumnPos {
pub name: String,
pub pos: usize,
}
/// The meaning of the timestamp number produced by data sources. This type
/// is not concerned with the source of the timestamp (like if the data came
/// from a Debezium consistency topic or a CDCv2 stream), instead only what the
/// timestamp number means.
///
/// Some variants here have attached data used to differentiate incomparable
/// instantiations. These attached data types should be expanded in the future
/// if we need to tell apart more kinds of sources.
#[derive(Clone, Debug, Ord, PartialOrd, Eq, PartialEq, Serialize, Deserialize, Hash)]
pub enum Timeline {
/// EpochMilliseconds means the timestamp is the number of milliseconds since
/// the Unix epoch.
EpochMilliseconds,
/// External means the timestamp comes from an external data source and we
/// don't know what the number means. The attached String is the source's name,
/// which will result in different sources being incomparable.
External(String),
/// User means the user has manually specified a timeline. The attached
/// String is specified by the user, allowing them to decide sources that are
/// joinable.
User(String),
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum SourceEnvelope {
/// If present, include the key columns as an output column of the source with the given properties.
None(KeyEnvelope),
Debezium(DebeziumEnvelope),
Upsert(KeyEnvelope),
CdcV2,
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct DebeziumEnvelope {
/// The column index containing the `before` row
pub before_idx: usize,
/// The column index containing the `after` row
pub after_idx: usize,
pub mode: DebeziumMode,
}
/// Ordered means we can trust Debezium high water marks
///
/// In standard operation, Debezium should always emit messages in position order, but
/// messages may be duplicated.
///
/// For example, this is a legal stream of Debezium event positions:
///
/// ```text
/// 1 2 3 2
/// ```
///
/// Note that `2` appears twice, but the *first* time it appeared it appeared in order.
/// Any position below the highest-ever seen position is guaranteed to be a duplicate,
/// and can be ignored.
///
/// Now consider this stream:
///
/// ```text
/// 1 3 2
/// ```
///
/// In this case, `2` is sent *out* of order, and if it is ignored we will miss important
/// state.
///
/// It is possible for users to do things with multiple databases and multiple Debezium
/// instances pointing at the same Kafka topic that mean that the Debezium guarantees do
/// not hold, in which case we are required to track individual messages, instead of just
/// the highest-ever-seen message.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum DebeziumMode {
/// Do not perform any deduplication
None,
/// Keep track of keys from upstream and discard retractions for new keys
Upsert,
/// We can trust high water mark
Ordered(DebeziumDedupProjection),
/// We need to store some piece of state for every message
Full(DebeziumDedupProjection),
FullInRange {
projection: DebeziumDedupProjection,
pad_start: Option<NaiveDateTime>,
start: NaiveDateTime,
end: NaiveDateTime,
},
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct DebeziumDedupProjection {
/// The column index containing the debezium source metadata
pub source_idx: usize,
/// The record index of the `source.snapshot` field
pub snapshot_idx: usize,
/// The upstream database specific fields
pub source_projection: DebeziumSourceProjection,
/// The column index containing the debezium transaction metadata
pub transaction_idx: usize,
/// The record index of the `transaction.total_order` field
pub total_order_idx: usize,
}
/// Debezium generates records that contain metadata about the upstream database. The structure of
/// this metadata depends on the type of connector used. This struct records the relevant indices
/// in the record, calculated during planning, so that the dataflow operator can unpack the
/// structure and extract the relevant information.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum DebeziumSourceProjection {
MySql {
file: usize,
pos: usize,
row: usize,
},
Postgres {
sequence: Option<usize>,
lsn: usize,
},
SqlServer {
change_lsn: usize,
event_serial_no: usize,
},
}
/// Computes the indices of the value's relation description that appear in the key.
///
/// Returns an error if it detects a common columns between the two relations that has the same
/// name but a different type, if a key column is missing from the value, and if the key relation
/// has a column with no name.
fn match_key_indices(
key_desc: &RelationDesc,
value_desc: &RelationDesc,
) -> anyhow::Result<Vec<usize>> {
let mut indices = Vec::new();
for (name, key_type) in key_desc.iter() {
let (index, value_type) = value_desc
.get_by_name(name)
.ok_or_else(|| anyhow!("Value schema missing primary key column: {}", name))?;
if key_type == value_type {
indices.push(index);
} else {
bail!(
"key and value column types do not match: key {:?} vs. value {:?}",
key_type,
value_type
);
}
}
Ok(indices)
}
impl SourceEnvelope {
/// Computes the output relation of this envelope when applied on top of the decoded key and
/// value relation desc
pub fn desc(
&self,
key_desc: Option<RelationDesc>,
value_desc: RelationDesc,
metadata_desc: RelationDesc,
) -> anyhow::Result<RelationDesc> {
Ok(match self {
SourceEnvelope::None(key_envelope) | SourceEnvelope::Upsert(key_envelope) => {
let key_desc = match key_desc {
Some(desc) => desc,
None => return Ok(value_desc.concat(metadata_desc)),
};
let keyed = match key_envelope {
KeyEnvelope::None => value_desc,
KeyEnvelope::Flattened => {
// Add the key columns as a key.
let key_indices = (0..key_desc.arity()).collect();
let key_desc = key_desc.with_key(key_indices);
key_desc.concat(value_desc)
}
KeyEnvelope::LegacyUpsert => {
let key_indices = (0..key_desc.arity()).collect();
let key_desc = key_desc.with_key(key_indices);
let names = (0..key_desc.arity()).map(|i| format!("key{}", i));
// Rename key columns to "keyN"
key_desc.with_names(names).concat(value_desc)
}
KeyEnvelope::Named(key_name) => {
let key_desc = {
// if the key has multiple objects, nest them as a record inside of a single name
if key_desc.arity() > 1 {
let key_type = key_desc.typ();
let key_as_record = RelationType::new(vec![ColumnType {
nullable: false,
scalar_type: ScalarType::Record {
fields: key_desc
.iter_names()
.zip(key_type.column_types.iter())
.map(|(name, ty)| (name.clone(), ty.clone()))
.collect(),
custom_oid: None,
custom_name: None,
},
}]);
RelationDesc::new(key_as_record, [key_name.to_string()])
} else {
key_desc.with_names([key_name.to_string()])
}
};
// In all cases the first column is the key
key_desc.with_key(vec![0]).concat(value_desc)
}
};
keyed.concat(metadata_desc)
}
SourceEnvelope::Debezium(envelope) => {
match &value_desc.typ().column_types[envelope.after_idx].scalar_type {
ScalarType::Record { fields, .. } => {
let mut desc = RelationDesc::from_names_and_types(fields.clone());
if let Some(key) = key_desc.map(|k| match_key_indices(&k, &desc)) {
desc = desc.with_key(key?);
}
match envelope.mode {
DebeziumMode::Upsert => desc.concat(metadata_desc),
_ => desc,
}
}
ty => bail!(
"Incorrect type for Debezium value, expected Record, got {:?}",
ty
),
}
}
SourceEnvelope::CdcV2 => {
// TODO: Validate that the value schema has an `updates` and `progress` column of
// the correct types
// CdcV2 row data are in a record in a record in a list
match &value_desc.typ().column_types[0].scalar_type {
ScalarType::List { element_type, .. } => match &**element_type {
ScalarType::Record { fields, .. } => {
// TODO maybe check this by name
match &fields[0].1.scalar_type {
ScalarType::Record { fields, .. } => {
RelationDesc::from_names_and_types(fields.clone())
}
ty => {
bail!("Unepxected type for MATERIALIZE envelope: {:?}", ty)
}
}
}
ty => bail!("Unepxected type for MATERIALIZE envelope: {:?}", ty),
},
ty => bail!("Unepxected type for MATERIALIZE envelope: {:?}", ty),
}
}
})
}
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct KafkaSourceConnector {
pub addrs: KafkaAddrs,
pub topic: String,
// Represents options specified by user when creating the source, e.g.
// security settings.
pub config_options: BTreeMap<String, String>,
// Map from partition -> starting offset
pub start_offsets: HashMap<i32, i64>,
pub group_id_prefix: Option<String>,
pub cluster_id: Uuid,
/// If present, include the timestamp as an output column of the source with the given name
pub include_timestamp: Option<IncludedColumnPos>,
/// If present, include the partition as an output column of the source with the given name.
pub include_partition: Option<IncludedColumnPos>,
/// If present, include the topic as an output column of the source with the given name.
pub include_topic: Option<IncludedColumnPos>,
/// If present, include the offset as an output column of the source with the given name.
pub include_offset: Option<IncludedColumnPos>,
}
/// Legacy logic included something like an offset into almost data streams
///
/// Eventually we will require `INCLUDE <metadata>` for everything.
pub fn provide_default_metadata(
envelope: &SourceEnvelope,
encoding: &encoding::DataEncoding,
) -> bool {
let is_avro = matches!(encoding, encoding::DataEncoding::Avro(_));
let is_stateless_dbz = match envelope {
SourceEnvelope::Debezium(DebeziumEnvelope {
mode: DebeziumMode::Upsert,
..
}) => false,
SourceEnvelope::Debezium(_) => true,
_ => false,
};
!is_avro && !is_stateless_dbz
}
#[derive(Clone, Copy, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum Compression {
Gzip,
None,
}
/// A source of updates for a relational collection.
///
/// A source contains enough information to instantiate a stream of changes,
/// as well as related metadata about the columns, their types, and properties
/// of the collection.
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub struct SourceDesc {
pub name: String,
pub connector: SourceConnector,
pub desc: RelationDesc,
}
#[derive(Clone, Debug, Serialize, Deserialize, Eq, PartialEq)]
pub enum SourceConnector {
External {
connector: ExternalSourceConnector,
encoding: encoding::SourceDataEncoding,
envelope: SourceEnvelope,
ts_frequency: Duration,
timeline: Timeline,
},
/// A local "source" is either fed by a local input handle, or by reading from a
/// `persisted_source()`. For non-persisted sources, values that are to be inserted
/// are sent from the coordinator and pushed into the handle on a worker.
///
/// For persisted sources, the coordinator only writes new values to a persistent
/// stream. These values will then "show up" here because we read from the same
/// persistent stream.
// TODO: We could split this up into a `Local` source, that is only fed by a local handle and a
// `LocalPersistenceSource` which is fed from a `persisted_source()`. But moving the
// persist_name from `SourceDesc` to here is already a huge simplification/clarification. The
// persist description on a `SourceDesc` is now purely used to signal that a source actively
// persists, while a `LocalPersistenceSource` is a source that happens to read from persistence
// but doesn't persist itself.
//
// That additional split seems like a bigger undertaking, though, because it also needs changes
// to the coordinator. And I don't know if I want to invest too much time there when I don't
// yet know how Tables will work in a post-ingestd world.
Local {
timeline: Timeline,
persisted_name: Option<String>,
},
}
impl SourceConnector {
/// Returns `true` if this connector yields input data (including
/// timestamps) that is stable across restarts. This is important for
/// exactly-once Sinks that need to ensure that the same data is written,
/// even when failures/restarts happen.
pub fn yields_stable_input(&self) -> bool {
if let SourceConnector::External { connector, .. } = self {
// Conservatively, set all Kafka, File, or AvroOcf sources as having stable inputs because
// we know they will be read in a known, repeatable offset order (modulo compaction for some Kafka sources).
match connector {
ExternalSourceConnector::Kafka(_)
| ExternalSourceConnector::File(_)
| ExternalSourceConnector::AvroOcf(_) => true,
// Currently, the Kinesis connector assigns "offsets" by counting the message in the order it was received
// and this order is not replayable across different reads of the same Kinesis stream.
ExternalSourceConnector::Kinesis(_) => false,
_ => false,
}
} else {
false
}
}
pub fn name(&self) -> &'static str {
match self {
SourceConnector::External { connector, .. } => connector.name(),
SourceConnector::Local { .. } => "local",
}
}
pub fn timeline(&self) -> Timeline {
match self {
SourceConnector::External { timeline, .. } => timeline.clone(),
SourceConnector::Local { timeline, .. } => timeline.clone(),
}
}
pub fn requires_single_materialization(&self) -> bool {
if let SourceConnector::External { connector, .. } = self {
connector.requires_single_materialization()
} else {
false
}
}
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum ExternalSourceConnector {
Kafka(KafkaSourceConnector),
Kinesis(KinesisSourceConnector),
File(FileSourceConnector),
AvroOcf(FileSourceConnector),
S3(S3SourceConnector),
Postgres(PostgresSourceConnector),
PubNub(PubNubSourceConnector),
}
impl ExternalSourceConnector {
/// Returns the name and type of each additional metadata column that
/// Materialize will automatically append to the source's inherent columns.
///
/// Presently, each source type exposes precisely one metadata column that
/// corresponds to some source-specific record counter. For example, file
/// sources use a line number, while Kafka sources use a topic offset.
///
/// The columns declared here must be kept in sync with the actual source
/// implementations that produce these columns.
pub fn metadata_columns(&self, include_defaults: bool) -> Vec<(&str, ColumnType)> {
let mut columns = Vec::new();
let default_col = |name| (name, ScalarType::Int64.nullable(false));
match self {
Self::Kafka(KafkaSourceConnector {
include_partition: part,
include_timestamp: time,
include_topic: topic,
include_offset: offset,
..
}) => {
let mut items = BTreeMap::new();
// put the offset at the end if necessary
if include_defaults && offset.is_none() {
items.insert(4, default_col("mz_offset"));
}
for (include, ty) in [
(offset, ScalarType::Int64),
(part, ScalarType::Int32),
(time, ScalarType::Timestamp),
(topic, ScalarType::String),
] {
if let Some(include) = include {
items.insert(include.pos + 1, (&include.name, ty.nullable(false)));
}
}
items.into_values().collect()
}
Self::File(_) => {
if include_defaults {
columns.push(default_col("mz_line_no"));
}
columns
}
Self::Kinesis(_) => {
if include_defaults {
columns.push(default_col("mz_offset"))
};
columns
}
Self::AvroOcf(_) => {
if include_defaults {
columns.push(default_col("mz_obj_no"))
};
columns
}
// TODO: should we include object key and possibly object-internal offset here?
Self::S3(_) => {
if include_defaults {
columns.push(default_col("mz_record"))
};
columns
}
Self::Postgres(_) => vec![],
Self::PubNub(_) => vec![],
}
}
// TODO(bwm): get rid of this when we no longer have the notion of default metadata
pub fn default_metadata_column_name(&self) -> Option<&str> {
match self {
ExternalSourceConnector::Kafka(_) => Some("mz_offset"),
ExternalSourceConnector::Kinesis(_) => Some("mz_offset"),
ExternalSourceConnector::File(_) => Some("mz_line_no"),
ExternalSourceConnector::AvroOcf(_) => Some("mz_obj_no"),
ExternalSourceConnector::S3(_) => Some("mz_record"),
ExternalSourceConnector::Postgres(_) => None,
ExternalSourceConnector::PubNub(_) => None,
}
}
pub fn metadata_column_types(&self, include_defaults: bool) -> Vec<IncludedColumnSource> {
match self {
ExternalSourceConnector::Kafka(KafkaSourceConnector {
include_partition: part,
include_timestamp: time,
include_topic: topic,
include_offset: offset,
..
}) => {
// create a sorted list of column types based on the order they were declared in sql
// TODO: should key be included in the sorted list? Breaking change, and it's
// already special (it commonly multiple columns embedded in it).
let mut items = BTreeMap::new();
if include_defaults {
items.insert(4, IncludedColumnSource::DefaultPosition);
}
for (include, ty) in [
(offset, IncludedColumnSource::Offset),
(part, IncludedColumnSource::Partition),
(time, IncludedColumnSource::Timestamp),
(topic, IncludedColumnSource::Topic),
] {
if let Some(include) = include {
items.insert(include.pos, ty);
}
}
items.into_values().collect()
}
ExternalSourceConnector::Kinesis(_)
| ExternalSourceConnector::File(_)
| ExternalSourceConnector::AvroOcf(_)
| ExternalSourceConnector::S3(_) => {
if include_defaults {
vec![IncludedColumnSource::DefaultPosition]
} else {
Vec::new()
}
}
ExternalSourceConnector::Postgres(_) | ExternalSourceConnector::PubNub(_) => {
Vec::new()
}
}
}
/// Returns the name of the external source connector.
pub fn name(&self) -> &'static str {
match self {
ExternalSourceConnector::Kafka(_) => "kafka",
ExternalSourceConnector::Kinesis(_) => "kinesis",
ExternalSourceConnector::File(_) => "file",
ExternalSourceConnector::AvroOcf(_) => "avro-ocf",
ExternalSourceConnector::S3(_) => "s3",
ExternalSourceConnector::Postgres(_) => "postgres",
ExternalSourceConnector::PubNub(_) => "pubnub",
}
}
/// Optionally returns the name of the upstream resource this source corresponds to.
/// (Currently only implemented for Kafka and Kinesis, to match old-style behavior
/// TODO: decide whether we want file paths and other upstream names to show up in metrics too.
pub fn upstream_name(&self) -> Option<&str> {
match self {
ExternalSourceConnector::Kafka(KafkaSourceConnector { topic, .. }) => {
Some(topic.as_str())
}
ExternalSourceConnector::Kinesis(KinesisSourceConnector {
stream_name, ..
}) => Some(stream_name.as_str()),
ExternalSourceConnector::File(_) => None,
ExternalSourceConnector::AvroOcf(_) => None,
ExternalSourceConnector::S3(_) => None,
ExternalSourceConnector::Postgres(_) => None,
ExternalSourceConnector::PubNub(_) => None,
}
}
pub fn requires_single_materialization(&self) -> bool {
match self {
ExternalSourceConnector::S3(c) => c.requires_single_materialization(),
ExternalSourceConnector::Postgres(_) => true,
ExternalSourceConnector::Kafka(_)
| ExternalSourceConnector::Kinesis(_)
| ExternalSourceConnector::File(_)
| ExternalSourceConnector::AvroOcf(_)
| ExternalSourceConnector::PubNub(_) => false,
}
}
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct KinesisSourceConnector {
pub stream_name: String,
pub aws: AwsConfig,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct FileSourceConnector {
pub path: PathBuf,
pub tail: bool,
pub compression: Compression,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct PostgresSourceConnector {
pub conn: String,
pub publication: String,
pub slot_name: String,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct PubNubSourceConnector {
pub subscribe_key: String,
pub channel: String,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct S3SourceConnector {
pub key_sources: Vec<S3KeySource>,
pub pattern: Option<Glob>,
pub aws: AwsConfig,
pub compression: Compression,
}
impl S3SourceConnector {
fn requires_single_materialization(&self) -> bool {
// SQS Notifications are not durable, multiple sources depending on them will get
// non-intersecting subsets of objects to read
self.key_sources
.iter()
.any(|s| matches!(s, S3KeySource::SqsNotifications { .. }))
}
}
/// A Source of Object Key names, the argument of the `DISCOVER OBJECTS` clause
#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub enum S3KeySource {
/// Scan the S3 Bucket to discover keys to download
Scan { bucket: String },
/// Load object keys based on the contents of an S3 Notifications channel
///
/// S3 notifications channels can be configured to go to SQS, which is the
/// only target we currently support.
SqsNotifications { queue: String },
}
/// A wrapper for [`Uri`] that implements [`Serialize`] and `Deserialize`.
#[derive(Debug, Clone, PartialEq, Eq, Hash, Serialize, Deserialize)]
pub struct SerdeUri(#[serde(with = "http_serde::uri")] pub Uri);
/// AWS configuration overrides for a source or sink.
///
/// This is a distinct type from any of the configuration types built into the
/// AWS SDK so that we can implement `Serialize` and `Deserialize`.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct AwsConfig {
/// AWS Credentials, or where to find them
pub credentials: AwsCredentials,
/// The AWS region to use.
///
/// Uses the default region (looking at env vars, config files, etc) if not provided.
pub region: Option<String>,
/// The AWS role to assume.
pub role: Option<AwsAssumeRole>,
/// The custom AWS endpoint to use, if any.
pub endpoint: Option<SerdeUri>,
}
/// AWS credentials for a source or sink.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum AwsCredentials {
/// Look for credentials using the [default credentials chain][credchain]
///
/// [credchain]: aws_config::default_provider::credentials::DefaultCredentialsChain
Default,
/// Load credentials using the given named profile
Profile { profile_name: String },
/// Use the enclosed static credentials
Static {
access_key_id: String,
secret_access_key: String,
session_token: Option<String>,
},
}
/// A role for Materialize to assume when performing AWS API calls.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct AwsAssumeRole {
/// The Amazon Resource Name of the role to assume.
pub arn: String,
/// The External ID for this customer Materialize provides during role assumption.
///
/// <https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_create_for-user_externalid.html>
pub external_id: Option<String>,
}
impl AwsConfig {
/// Loads the AWS SDK configuration object from the environment, then
/// applies the overrides from this object.
pub async fn load(&self) -> mz_aws_util::config::AwsConfig {
use aws_config::default_provider::credentials::DefaultCredentialsChain;
use aws_config::default_provider::region::DefaultRegionChain;
use aws_config::sts::AssumeRoleProvider;
use aws_smithy_http::endpoint::Endpoint;
use aws_types::credentials::SharedCredentialsProvider;
use aws_types::region::Region;
let region = match &self.region {
Some(region) => Some(Region::new(region.clone())),
_ => {
let mut rc = DefaultRegionChain::builder();
if let AwsCredentials::Profile { profile_name } = &self.credentials {
rc = rc.profile_name(profile_name);
}
// This doesn't work for profiles which do not provide their own region but
// instead specify a source_profile that does have a region.
//
// https://github.com/awslabs/aws-sdk-rust/issues/443
rc.build().region().await
}
};
let mut cred_provider = match &self.credentials {
AwsCredentials::Default => SharedCredentialsProvider::new(
DefaultCredentialsChain::builder()
.region(region.clone())
.build()
.await,
),
AwsCredentials::Profile { profile_name } => SharedCredentialsProvider::new(
DefaultCredentialsChain::builder()
.profile_name(profile_name)
.region(region.clone())
.build()
.await,
),
AwsCredentials::Static {
access_key_id,
secret_access_key,
session_token,
} => SharedCredentialsProvider::new(aws_types::Credentials::from_keys(
access_key_id,
secret_access_key,
session_token.clone(),
)),
};
if let Some(AwsAssumeRole { arn, external_id }) = &self.role {
let mut role = AssumeRoleProvider::builder(arn).session_name("materialized");
// This affects which region to perform STS on, not where
// anything else happens.
if let Some(region) = ®ion {
role = role.region(region.clone());
}
if let Some(external_id) = &external_id {
role = role.external_id(external_id);
}
cred_provider = SharedCredentialsProvider::new(role.build(cred_provider));
}
let loader = aws_config::from_env()
.region(region)
.credentials_provider(cred_provider);
let mut config = mz_aws_util::config::AwsConfig::from_loader(loader).await;
if let Some(endpoint) = &self.endpoint {
config.set_endpoint(Endpoint::immutable(endpoint.0.clone()));
}
config
}
}
}
/// Types and traits related to reporting changing collections out of `dataflow`.
pub mod sinks {
use std::collections::BTreeMap;
use std::path::PathBuf;
use serde::{Deserialize, Serialize};
use timely::progress::frontier::Antichain;
use url::Url;
use expr::GlobalId;
use kafka_util::KafkaAddrs;
use repr::{RelationDesc, Timestamp};
/// A sink for updates to a relational collection.
#[derive(Clone, Debug, Serialize, Deserialize)]
pub struct SinkDesc {
pub from: GlobalId,
pub from_desc: RelationDesc,
pub connector: SinkConnector,
pub envelope: Option<SinkEnvelope>,
pub as_of: SinkAsOf,
}
#[derive(Copy, Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum SinkEnvelope {
Debezium,
Upsert,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct SinkAsOf {
pub frontier: Antichain<Timestamp>,
pub strict: bool,
}
#[derive(Clone, Debug, Serialize, Deserialize)]
pub enum SinkConnector {
Kafka(KafkaSinkConnector),
Tail(TailSinkConnector),
AvroOcf(AvroOcfSinkConnector),
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct KafkaSinkConsistencyConnector {
pub topic: String,
pub schema_id: i32,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct KafkaSinkConnector {
pub addrs: KafkaAddrs,
pub topic: String,
pub topic_prefix: String,
pub key_desc_and_indices: Option<(RelationDesc, Vec<usize>)>,
pub relation_key_indices: Option<Vec<usize>>,
pub value_desc: RelationDesc,
pub published_schema_info: Option<PublishedSchemaInfo>,
pub consistency: Option<KafkaSinkConsistencyConnector>,
pub exactly_once: bool,
// Source dependencies for exactly-once sinks.
pub transitive_source_dependencies: Vec<GlobalId>,
// Maximum number of records the sink will attempt to send each time it is
// invoked
pub fuel: usize,
pub config_options: BTreeMap<String, String>,
}
/// TODO(JLDLaughlin): Documentation.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct PublishedSchemaInfo {
pub key_schema_id: Option<i32>,
pub value_schema_id: i32,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct AvroOcfSinkConnector {
pub value_desc: RelationDesc,
pub path: PathBuf,
}
impl SinkConnector {
/// Returns the name of the sink connector.
pub fn name(&self) -> &'static str {
match self {
SinkConnector::AvroOcf(_) => "avro-ocf",
SinkConnector::Kafka(_) => "kafka",
SinkConnector::Tail(_) => "tail",
}
}
/// Returns `true` if this sink requires sources to block timestamp binding
/// compaction until all sinks that depend on a given source have finished
/// writing out that timestamp.
///
/// To achieve that, each sink will hold a `AntichainToken` for all of
/// the sources it depends on, and will advance all of its source
/// dependencies' compaction frontiers as it completes writes.
///
/// Sinks that do need to hold back compaction need to insert an
/// [`Antichain`] into `StorageState::sink_write_frontiers` that they update
/// in order to advance the frontier that holds back upstream compaction
/// of timestamp bindings.
///
/// See also [`transitive_source_dependencies`](SinkConnector::transitive_source_dependencies).
pub fn requires_source_compaction_holdback(&self) -> bool {
match self {
SinkConnector::Kafka(k) => k.exactly_once,
SinkConnector::AvroOcf(_) => false,
SinkConnector::Tail(_) => false,
}
}
/// Returns the [`GlobalIds`](GlobalId) of the transitive sources of this
/// sink.
pub fn transitive_source_dependencies(&self) -> &[GlobalId] {
match self {
SinkConnector::Kafka(k) => &k.transitive_source_dependencies,
SinkConnector::AvroOcf(_) => &[],
SinkConnector::Tail(_) => &[],
}
}
}
#[derive(Default, Clone, Debug, Serialize, Deserialize)]
pub struct TailSinkConnector {}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum SinkConnectorBuilder {
Kafka(KafkaSinkConnectorBuilder),
AvroOcf(AvroOcfSinkConnectorBuilder),
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct AvroOcfSinkConnectorBuilder {
pub path: PathBuf,
pub file_name_suffix: String,
pub value_desc: RelationDesc,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub struct KafkaSinkConnectorBuilder {
pub broker_addrs: KafkaAddrs,
pub format: KafkaSinkFormat,
/// A natural key of the sinked relation (view or source).
pub relation_key_indices: Option<Vec<usize>>,
/// The user-specified key for the sink.
pub key_desc_and_indices: Option<(RelationDesc, Vec<usize>)>,
pub value_desc: RelationDesc,
pub topic_prefix: String,
pub consistency_topic_prefix: Option<String>,
pub consistency_format: Option<KafkaSinkFormat>,
pub topic_suffix_nonce: String,
pub partition_count: i32,
pub replication_factor: i32,
pub fuel: usize,
pub config_options: BTreeMap<String, String>,
// Forces the sink to always write to the same topic across restarts instead
// of picking a new topic each time.
pub reuse_topic: bool,
// Source dependencies for exactly-once sinks.
pub transitive_source_dependencies: Vec<GlobalId>,
pub retention: KafkaSinkConnectorRetention,
}
#[derive(Clone, Debug, Default, Eq, PartialEq, Serialize, Deserialize)]
pub struct KafkaSinkConnectorRetention {
pub retention_ms: Option<i64>,
pub retention_bytes: Option<i64>,
}
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize)]
pub enum KafkaSinkFormat {
Avro {
schema_registry_url: Url,
key_schema: Option<String>,
value_schema: String,
ccsr_config: ccsr::ClientConfig,
},
Json,
}
}
/// An index storing processed updates so they can be queried
/// or reused in other computations
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize, Hash)]
pub struct IndexDesc {
/// Identity of the collection the index is on.
pub on_id: GlobalId,
/// Expressions to be arranged, in order of decreasing primacy.
pub key: Vec<MirScalarExpr>,
}
// TODO: change contract to ensure that the operator is always applied to
// streams of rows
/// In-place restrictions that can be made to rows.
///
/// These fields indicate *optional* information that may applied to
/// streams of rows. Any row that does not satisfy all predicates may
/// be discarded, and any column not listed in the projection may be
/// replaced by a default value.
///
/// The intended order of operations is that the predicates are first
/// applied, and columns not in projection can then be overwritten with
/// default values. This allows the projection to avoid capturing columns
/// used by the predicates but not otherwise required.
#[derive(Clone, Debug, Eq, PartialEq, Serialize, Deserialize, Hash)]
pub struct LinearOperator {
/// Rows that do not pass all predicates may be discarded.
pub predicates: Vec<MirScalarExpr>,
/// Columns not present in `projection` may be replaced with
/// default values.
pub projection: Vec<usize>,
}
impl LinearOperator {
/// Reports whether this linear operator is trivial when applied to an
/// input of the specified arity.
pub fn is_trivial(&self, arity: usize) -> bool {
self.predicates.is_empty() && self.projection.iter().copied().eq(0..arity)
}
}