Search in sources :

Example 1 with ReadingMetadataSpec

use of org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec in project flink by apache.

the class PushProjectIntoTableSourceScanRule method matches.

@Override
public boolean matches(RelOptRuleCall call) {
    final LogicalTableScan scan = call.rel(1);
    final TableSourceTable sourceTable = scan.getTable().unwrap(TableSourceTable.class);
    if (sourceTable == null) {
        return false;
    }
    final DynamicTableSource source = sourceTable.tableSource();
    // The source supports projection push-down.
    if (supportsProjectionPushDown(source)) {
        return Arrays.stream(sourceTable.abilitySpecs()).noneMatch(spec -> spec instanceof ProjectPushDownSpec);
    }
    // (for physical columns) is not supported.
    if (supportsMetadata(source)) {
        if (Arrays.stream(sourceTable.abilitySpecs()).anyMatch(spec -> spec instanceof ReadingMetadataSpec)) {
            return false;
        }
        return ((SupportsReadingMetadata) source).supportsMetadataProjection();
    }
    return false;
}
Also used : ProjectPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.ProjectPushDownSpec) SupportsReadingMetadata(org.apache.flink.table.connector.source.abilities.SupportsReadingMetadata) TableSourceTable(org.apache.flink.table.planner.plan.schema.TableSourceTable) ReadingMetadataSpec(org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec) LogicalTableScan(org.apache.calcite.rel.logical.LogicalTableScan) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource)

Example 2 with ReadingMetadataSpec

use of org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec in project flink by apache.

the class PushProjectIntoTableSourceScanRule method performPushDown.

private RowType performPushDown(TableSourceTable source, NestedSchema projectedSchema, RowType producedType, List<SourceAbilitySpec> abilitySpecs) {
    final int numPhysicalColumns;
    final List<NestedColumn> projectedMetadataColumns;
    if (supportsMetadata(source.tableSource())) {
        final List<String> declaredMetadataKeys = createRequiredMetadataKeys(source.contextResolvedTable().getResolvedSchema(), source.tableSource());
        numPhysicalColumns = producedType.getFieldCount() - declaredMetadataKeys.size();
        projectedMetadataColumns = IntStream.range(0, declaredMetadataKeys.size()).mapToObj(i -> producedType.getFieldNames().get(numPhysicalColumns + i)).map(fieldName -> projectedSchema.columns().get(fieldName)).filter(Objects::nonNull).collect(Collectors.toList());
    } else {
        numPhysicalColumns = producedType.getFieldCount();
        projectedMetadataColumns = Collections.emptyList();
    }
    final int[][] physicalProjections;
    if (supportsProjectionPushDown(source.tableSource())) {
        projectedMetadataColumns.forEach(metaColumn -> projectedSchema.columns().remove(metaColumn.name()));
        physicalProjections = NestedProjectionUtil.convertToIndexArray(projectedSchema);
        projectedMetadataColumns.forEach(metaColumn -> projectedSchema.columns().put(metaColumn.name(), metaColumn));
    } else {
        physicalProjections = IntStream.range(0, numPhysicalColumns).mapToObj(columnIndex -> new int[] { columnIndex }).toArray(int[][]::new);
    }
    final int[][] projectedFields = Stream.concat(Stream.of(physicalProjections), projectedMetadataColumns.stream().map(NestedColumn::indexInOriginSchema).map(columnIndex -> new int[] { columnIndex })).toArray(int[][]::new);
    int newIndex = physicalProjections.length;
    for (NestedColumn metaColumn : projectedMetadataColumns) {
        metaColumn.setIndexOfLeafInNewSchema(newIndex++);
    }
    if (supportsProjectionPushDown(source.tableSource())) {
        final RowType projectedPhysicalType = (RowType) Projection.of(physicalProjections).project(producedType);
        abilitySpecs.add(new ProjectPushDownSpec(physicalProjections, projectedPhysicalType));
    }
    final RowType newProducedType = (RowType) Projection.of(projectedFields).project(producedType);
    if (supportsMetadata(source.tableSource())) {
        final List<String> projectedMetadataKeys = projectedMetadataColumns.stream().map(NestedColumn::name).collect(Collectors.toList());
        abilitySpecs.add(new ReadingMetadataSpec(projectedMetadataKeys, newProducedType));
    }
    return newProducedType;
}
Also used : IntStream(java.util.stream.IntStream) NestedProjectionUtil(org.apache.flink.table.planner.plan.utils.NestedProjectionUtil) Arrays(java.util.Arrays) ShortcutUtils.unwrapTypeFactory(org.apache.flink.table.planner.utils.ShortcutUtils.unwrapTypeFactory) SourceAbilityContext(org.apache.flink.table.planner.plan.abilities.source.SourceAbilityContext) Column(org.apache.flink.table.catalog.Column) ResolvedSchema(org.apache.flink.table.catalog.ResolvedSchema) RexNodeExtractor(org.apache.flink.table.planner.plan.utils.RexNodeExtractor) FlinkTypeFactory(org.apache.flink.table.planner.calcite.FlinkTypeFactory) RowType(org.apache.flink.table.types.logical.RowType) SupportsProjectionPushDown(org.apache.flink.table.connector.source.abilities.SupportsProjectionPushDown) ArrayList(java.util.ArrayList) RexNode(org.apache.calcite.rex.RexNode) NestedSchema(org.apache.flink.table.planner.plan.utils.NestedSchema) Projection(org.apache.flink.table.connector.Projection) ProjectRemoveRule(org.apache.calcite.rel.rules.ProjectRemoveRule) DynamicSourceUtils.createProducedType(org.apache.flink.table.planner.connectors.DynamicSourceUtils.createProducedType) RelDataType(org.apache.calcite.rel.type.RelDataType) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) TableConfig(org.apache.flink.table.api.TableConfig) LogicalProject(org.apache.calcite.rel.logical.LogicalProject) ProjectPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.ProjectPushDownSpec) TableException(org.apache.flink.table.api.TableException) ShortcutUtils.unwrapContext(org.apache.flink.table.planner.utils.ShortcutUtils.unwrapContext) RelRule(org.apache.calcite.plan.RelRule) NestedColumn(org.apache.flink.table.planner.plan.utils.NestedColumn) Collectors(java.util.stream.Collectors) DynamicSourceUtils.createRequiredMetadataKeys(org.apache.flink.table.planner.connectors.DynamicSourceUtils.createRequiredMetadataKeys) SourceAbilitySpec(org.apache.flink.table.planner.plan.abilities.source.SourceAbilitySpec) TableSourceTable(org.apache.flink.table.planner.plan.schema.TableSourceTable) RelOptRuleCall(org.apache.calcite.plan.RelOptRuleCall) RexInputRef(org.apache.calcite.rex.RexInputRef) Objects(java.util.Objects) DynamicSourceUtils(org.apache.flink.table.planner.connectors.DynamicSourceUtils) RelOptRule(org.apache.calcite.plan.RelOptRule) List(java.util.List) Stream(java.util.stream.Stream) UniqueConstraint(org.apache.flink.table.catalog.UniqueConstraint) SupportsReadingMetadata(org.apache.flink.table.connector.source.abilities.SupportsReadingMetadata) ReadingMetadataSpec(org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec) Internal(org.apache.flink.annotation.Internal) Collections(java.util.Collections) LogicalTableScan(org.apache.calcite.rel.logical.LogicalTableScan) ProjectPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.ProjectPushDownSpec) Objects(java.util.Objects) NestedColumn(org.apache.flink.table.planner.plan.utils.NestedColumn) RowType(org.apache.flink.table.types.logical.RowType) ReadingMetadataSpec(org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec) UniqueConstraint(org.apache.flink.table.catalog.UniqueConstraint)

Example 3 with ReadingMetadataSpec

use of org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec in project flink by apache.

the class DynamicTableSourceSpecSerdeTest method testDynamicTableSinkSpecSerde.

public static Stream<DynamicTableSourceSpec> testDynamicTableSinkSpecSerde() {
    Map<String, String> options1 = new HashMap<>();
    options1.put("connector", FileSystemTableFactory.IDENTIFIER);
    options1.put("format", TestCsvFormatFactory.IDENTIFIER);
    options1.put("path", "/tmp");
    final ResolvedSchema resolvedSchema1 = new ResolvedSchema(Collections.singletonList(Column.physical("a", DataTypes.BIGINT())), Collections.emptyList(), null);
    final CatalogTable catalogTable1 = CatalogTable.of(Schema.newBuilder().fromResolvedSchema(resolvedSchema1).build(), null, Collections.emptyList(), options1);
    DynamicTableSourceSpec spec1 = new DynamicTableSourceSpec(ContextResolvedTable.temporary(ObjectIdentifier.of(DEFAULT_BUILTIN_CATALOG, DEFAULT_BUILTIN_DATABASE, "MyTable"), new ResolvedCatalogTable(catalogTable1, resolvedSchema1)), null);
    Map<String, String> options2 = new HashMap<>();
    options2.put("connector", TestValuesTableFactory.IDENTIFIER);
    options2.put("disable-lookup", "true");
    options2.put("enable-watermark-push-down", "true");
    options2.put("filterable-fields", "b");
    options2.put("bounded", "false");
    options2.put("readable-metadata", "m1:INT, m2:STRING");
    final ResolvedSchema resolvedSchema2 = new ResolvedSchema(Arrays.asList(Column.physical("a", DataTypes.BIGINT()), Column.physical("b", DataTypes.INT()), Column.physical("c", DataTypes.STRING()), Column.physical("p", DataTypes.STRING()), Column.metadata("m1", DataTypes.INT(), null, false), Column.metadata("m2", DataTypes.STRING(), null, false), Column.physical("ts", DataTypes.TIMESTAMP(3))), Collections.emptyList(), null);
    final CatalogTable catalogTable2 = CatalogTable.of(Schema.newBuilder().fromResolvedSchema(resolvedSchema2).build(), null, Collections.emptyList(), options2);
    FlinkTypeFactory factory = FlinkTypeFactory.INSTANCE();
    RexBuilder rexBuilder = new RexBuilder(factory);
    DynamicTableSourceSpec spec2 = new DynamicTableSourceSpec(ContextResolvedTable.temporary(ObjectIdentifier.of(DEFAULT_BUILTIN_CATALOG, DEFAULT_BUILTIN_DATABASE, "MyTable"), new ResolvedCatalogTable(catalogTable2, resolvedSchema2)), Arrays.asList(new ProjectPushDownSpec(new int[][] { { 0 }, { 1 }, { 4 }, { 6 } }, RowType.of(new LogicalType[] { new BigIntType(), new IntType(), new IntType(), new TimestampType(3) }, new String[] { "a", "b", "m1", "ts" })), new ReadingMetadataSpec(Arrays.asList("m1", "m2"), RowType.of(new LogicalType[] { new BigIntType(), new IntType(), new IntType(), new TimestampType(3) }, new String[] { "a", "b", "m1", "ts" })), new FilterPushDownSpec(Collections.singletonList(// b >= 10
    rexBuilder.makeCall(SqlStdOperatorTable.GREATER_THAN_OR_EQUAL, rexBuilder.makeInputRef(factory.createSqlType(SqlTypeName.INTEGER), 1), rexBuilder.makeExactLiteral(new BigDecimal(10))))), new WatermarkPushDownSpec(rexBuilder.makeCall(SqlStdOperatorTable.MINUS, rexBuilder.makeInputRef(factory.createSqlType(SqlTypeName.TIMESTAMP, 3), 3), rexBuilder.makeIntervalLiteral(BigDecimal.valueOf(1000), new SqlIntervalQualifier(TimeUnit.SECOND, 2, TimeUnit.SECOND, 6, SqlParserPos.ZERO))), 5000, RowType.of(new BigIntType(), new IntType(), new IntType(), new TimestampType(false, TimestampKind.ROWTIME, 3))), new SourceWatermarkSpec(true, RowType.of(new BigIntType(), new IntType(), new IntType(), new TimestampType(false, TimestampKind.ROWTIME, 3))), new LimitPushDownSpec(100), new PartitionPushDownSpec(Arrays.asList(new HashMap<String, String>() {

        {
            put("p", "A");
        }
    }, new HashMap<String, String>() {

        {
            put("p", "B");
        }
    }))));
    return Stream.of(spec1, spec2);
}
Also used : WatermarkPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.WatermarkPushDownSpec) HashMap(java.util.HashMap) ProjectPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.ProjectPushDownSpec) SqlIntervalQualifier(org.apache.calcite.sql.SqlIntervalQualifier) DynamicTableSourceSpec(org.apache.flink.table.planner.plan.nodes.exec.spec.DynamicTableSourceSpec) LogicalType(org.apache.flink.table.types.logical.LogicalType) BigIntType(org.apache.flink.table.types.logical.BigIntType) CatalogTable(org.apache.flink.table.catalog.CatalogTable) ResolvedCatalogTable(org.apache.flink.table.catalog.ResolvedCatalogTable) ReadingMetadataSpec(org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec) BigDecimal(java.math.BigDecimal) IntType(org.apache.flink.table.types.logical.IntType) BigIntType(org.apache.flink.table.types.logical.BigIntType) SourceWatermarkSpec(org.apache.flink.table.planner.plan.abilities.source.SourceWatermarkSpec) LimitPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.LimitPushDownSpec) PartitionPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.PartitionPushDownSpec) FilterPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.FilterPushDownSpec) ResolvedCatalogTable(org.apache.flink.table.catalog.ResolvedCatalogTable) FlinkTypeFactory(org.apache.flink.table.planner.calcite.FlinkTypeFactory) RexBuilder(org.apache.calcite.rex.RexBuilder) TimestampType(org.apache.flink.table.types.logical.TimestampType) ResolvedSchema(org.apache.flink.table.catalog.ResolvedSchema)

Aggregations

ProjectPushDownSpec (org.apache.flink.table.planner.plan.abilities.source.ProjectPushDownSpec)3 ReadingMetadataSpec (org.apache.flink.table.planner.plan.abilities.source.ReadingMetadataSpec)3 LogicalTableScan (org.apache.calcite.rel.logical.LogicalTableScan)2 ResolvedSchema (org.apache.flink.table.catalog.ResolvedSchema)2 DynamicTableSource (org.apache.flink.table.connector.source.DynamicTableSource)2 SupportsReadingMetadata (org.apache.flink.table.connector.source.abilities.SupportsReadingMetadata)2 FlinkTypeFactory (org.apache.flink.table.planner.calcite.FlinkTypeFactory)2 TableSourceTable (org.apache.flink.table.planner.plan.schema.TableSourceTable)2 BigDecimal (java.math.BigDecimal)1 ArrayList (java.util.ArrayList)1 Arrays (java.util.Arrays)1 Collections (java.util.Collections)1 HashMap (java.util.HashMap)1 List (java.util.List)1 Objects (java.util.Objects)1 Collectors (java.util.stream.Collectors)1 IntStream (java.util.stream.IntStream)1 Stream (java.util.stream.Stream)1 RelOptRule (org.apache.calcite.plan.RelOptRule)1 RelOptRuleCall (org.apache.calcite.plan.RelOptRuleCall)1