Search in sources :

Example 21 with DynamicTableSource

use of org.apache.flink.table.connector.source.DynamicTableSource in project flink by apache.

the class FactoryUtilTest method testAllOptions.

@Test
public void testAllOptions() {
    final Map<String, String> options = createAllOptions();
    final DynamicTableSource actualSource = createTableSource(SCHEMA, options);
    final DynamicTableSource expectedSource = new DynamicTableSourceMock("MyTarget", null, new DecodingFormatMock(",", false), new DecodingFormatMock("|", true));
    assertThat(actualSource).isEqualTo(expectedSource);
    final DynamicTableSink actualSink = createTableSink(SCHEMA, options);
    final DynamicTableSink expectedSink = new DynamicTableSinkMock("MyTarget", 1000L, new EncodingFormatMock(","), new EncodingFormatMock("|"));
    assertThat(actualSink).isEqualTo(expectedSink);
}
Also used : EncodingFormatMock(org.apache.flink.table.factories.TestFormatFactory.EncodingFormatMock) DecodingFormatMock(org.apache.flink.table.factories.TestFormatFactory.DecodingFormatMock) DynamicTableSink(org.apache.flink.table.connector.sink.DynamicTableSink) DynamicTableSourceMock(org.apache.flink.table.factories.TestDynamicTableFactory.DynamicTableSourceMock) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) DynamicTableSinkMock(org.apache.flink.table.factories.TestDynamicTableFactory.DynamicTableSinkMock) Test(org.junit.jupiter.api.Test)

Example 22 with DynamicTableSource

use of org.apache.flink.table.connector.source.DynamicTableSource in project flink by apache.

the class FactoryUtilTest method testManagedConnector.

@Test
public void testManagedConnector() {
    final Map<String, String> options = createAllOptions();
    options.remove("connector");
    final DynamicTableSource actualSource = createTableSource(SCHEMA, options);
    assertThat(actualSource).isExactlyInstanceOf(TestManagedTableSource.class);
}
Also used : DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) Test(org.junit.jupiter.api.Test)

Example 23 with DynamicTableSource

use of org.apache.flink.table.connector.source.DynamicTableSource in project flink by apache.

the class DataGenTableSourceFactoryTest method testSequenceCheckpointRestore.

@Test
public void testSequenceCheckpointRestore() throws Exception {
    DescriptorProperties descriptor = new DescriptorProperties();
    descriptor.putString(FactoryUtil.CONNECTOR.key(), "datagen");
    descriptor.putString(DataGenConnectorOptionsUtil.FIELDS + ".f0." + DataGenConnectorOptionsUtil.KIND, DataGenConnectorOptionsUtil.SEQUENCE);
    descriptor.putLong(DataGenConnectorOptionsUtil.FIELDS + ".f0." + DataGenConnectorOptionsUtil.START, 0);
    descriptor.putLong(DataGenConnectorOptionsUtil.FIELDS + ".f0." + DataGenConnectorOptionsUtil.END, 100);
    DynamicTableSource dynamicTableSource = createTableSource(ResolvedSchema.of(Column.physical("f0", DataTypes.BIGINT())), descriptor.asMap());
    DataGenTableSource dataGenTableSource = (DataGenTableSource) dynamicTableSource;
    DataGeneratorSource<RowData> source = dataGenTableSource.createSource();
    final int initElement = 0;
    final int maxElement = 100;
    final Set<RowData> expectedOutput = new HashSet<>();
    for (long i = initElement; i <= maxElement; i++) {
        expectedOutput.add(GenericRowData.of(i));
    }
    DataGeneratorSourceTest.innerTestDataGenCheckpointRestore(() -> {
        try {
            return InstantiationUtil.clone(source);
        } catch (IOException | ClassNotFoundException e) {
            throw new RuntimeException(e);
        }
    }, expectedOutput);
}
Also used : DescriptorProperties(org.apache.flink.table.descriptors.DescriptorProperties) IOException(java.io.IOException) GenericRowData(org.apache.flink.table.data.GenericRowData) RowData(org.apache.flink.table.data.RowData) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) DataGenTableSource(org.apache.flink.connector.datagen.table.DataGenTableSource) HashSet(java.util.HashSet) DataGeneratorSourceTest(org.apache.flink.streaming.api.functions.source.datagen.DataGeneratorSourceTest) Test(org.junit.Test)

Example 24 with DynamicTableSource

use of org.apache.flink.table.connector.source.DynamicTableSource in project flink by apache.

the class PushPartitionIntoTableSourceScanRule method onMatch.

@Override
public void onMatch(RelOptRuleCall call) {
    Filter filter = call.rel(0);
    LogicalTableScan scan = call.rel(1);
    TableSourceTable tableSourceTable = scan.getTable().unwrap(TableSourceTable.class);
    RelDataType inputFieldTypes = filter.getInput().getRowType();
    List<String> inputFieldNames = inputFieldTypes.getFieldNames();
    List<String> partitionFieldNames = tableSourceTable.contextResolvedTable().<ResolvedCatalogTable>getResolvedTable().getPartitionKeys();
    // extract partition predicates
    RelBuilder relBuilder = call.builder();
    RexBuilder rexBuilder = relBuilder.getRexBuilder();
    Tuple2<Seq<RexNode>, Seq<RexNode>> allPredicates = RexNodeExtractor.extractPartitionPredicateList(filter.getCondition(), FlinkRelOptUtil.getMaxCnfNodeCount(scan), inputFieldNames.toArray(new String[0]), rexBuilder, partitionFieldNames.toArray(new String[0]));
    RexNode partitionPredicate = RexUtil.composeConjunction(rexBuilder, JavaConversions.seqAsJavaList(allPredicates._1));
    if (partitionPredicate.isAlwaysTrue()) {
        return;
    }
    // build pruner
    LogicalType[] partitionFieldTypes = partitionFieldNames.stream().map(name -> {
        int index = inputFieldNames.indexOf(name);
        if (index < 0) {
            throw new TableException(String.format("Partitioned key '%s' isn't found in input columns. " + "Validator should have checked that.", name));
        }
        return inputFieldTypes.getFieldList().get(index).getType();
    }).map(FlinkTypeFactory::toLogicalType).toArray(LogicalType[]::new);
    RexNode finalPartitionPredicate = adjustPartitionPredicate(inputFieldNames, partitionFieldNames, partitionPredicate);
    FlinkContext context = ShortcutUtils.unwrapContext(scan);
    Function<List<Map<String, String>>, List<Map<String, String>>> defaultPruner = partitions -> PartitionPruner.prunePartitions(context.getTableConfig(), partitionFieldNames.toArray(new String[0]), partitionFieldTypes, partitions, finalPartitionPredicate);
    // prune partitions
    List<Map<String, String>> remainingPartitions = readPartitionsAndPrune(rexBuilder, context, tableSourceTable, defaultPruner, allPredicates._1(), inputFieldNames);
    // apply push down
    DynamicTableSource dynamicTableSource = tableSourceTable.tableSource().copy();
    PartitionPushDownSpec partitionPushDownSpec = new PartitionPushDownSpec(remainingPartitions);
    partitionPushDownSpec.apply(dynamicTableSource, SourceAbilityContext.from(scan));
    // build new statistic
    TableStats newTableStat = null;
    if (tableSourceTable.contextResolvedTable().isPermanent()) {
        ObjectIdentifier identifier = tableSourceTable.contextResolvedTable().getIdentifier();
        ObjectPath tablePath = identifier.toObjectPath();
        Catalog catalog = tableSourceTable.contextResolvedTable().getCatalog().get();
        for (Map<String, String> partition : remainingPartitions) {
            Optional<TableStats> partitionStats = getPartitionStats(catalog, tablePath, partition);
            if (!partitionStats.isPresent()) {
                // clear all information before
                newTableStat = null;
                break;
            } else {
                newTableStat = newTableStat == null ? partitionStats.get() : newTableStat.merge(partitionStats.get());
            }
        }
    }
    FlinkStatistic newStatistic = FlinkStatistic.builder().statistic(tableSourceTable.getStatistic()).tableStats(newTableStat).build();
    TableSourceTable newTableSourceTable = tableSourceTable.copy(dynamicTableSource, newStatistic, new SourceAbilitySpec[] { partitionPushDownSpec });
    LogicalTableScan newScan = LogicalTableScan.create(scan.getCluster(), newTableSourceTable, scan.getHints());
    // transform to new node
    RexNode nonPartitionPredicate = RexUtil.composeConjunction(rexBuilder, JavaConversions.seqAsJavaList(allPredicates._2()));
    if (nonPartitionPredicate.isAlwaysTrue()) {
        call.transformTo(newScan);
    } else {
        Filter newFilter = filter.copy(filter.getTraitSet(), newScan, nonPartitionPredicate);
        call.transformTo(newFilter);
    }
}
Also used : CatalogColumnStatistics(org.apache.flink.table.catalog.stats.CatalogColumnStatistics) Arrays(java.util.Arrays) SourceAbilityContext(org.apache.flink.table.planner.plan.abilities.source.SourceAbilityContext) PartitionNotExistException(org.apache.flink.table.catalog.exceptions.PartitionNotExistException) CatalogTable(org.apache.flink.table.catalog.CatalogTable) ShortcutUtils(org.apache.flink.table.planner.utils.ShortcutUtils) SupportsPartitionPushDown(org.apache.flink.table.connector.source.abilities.SupportsPartitionPushDown) FlinkTypeFactory(org.apache.flink.table.planner.calcite.FlinkTypeFactory) RexUtil(org.apache.calcite.rex.RexUtil) RexNode(org.apache.calcite.rex.RexNode) RelBuilder(org.apache.calcite.tools.RelBuilder) ResolvedExpression(org.apache.flink.table.expressions.ResolvedExpression) Map(java.util.Map) TableStats(org.apache.flink.table.plan.stats.TableStats) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) PartitionPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.PartitionPushDownSpec) TimeZone(java.util.TimeZone) Seq(scala.collection.Seq) FlinkContext(org.apache.flink.table.planner.calcite.FlinkContext) Tuple2(scala.Tuple2) Collectors(java.util.stream.Collectors) SourceAbilitySpec(org.apache.flink.table.planner.plan.abilities.source.SourceAbilitySpec) RexInputRef(org.apache.calcite.rex.RexInputRef) List(java.util.List) TableNotPartitionedException(org.apache.flink.table.catalog.exceptions.TableNotPartitionedException) LogicalType(org.apache.flink.table.types.logical.LogicalType) Optional(java.util.Optional) RexNodeToExpressionConverter(org.apache.flink.table.planner.plan.utils.RexNodeToExpressionConverter) LogicalTableScan(org.apache.calcite.rel.logical.LogicalTableScan) ObjectIdentifier(org.apache.flink.table.catalog.ObjectIdentifier) CatalogTableStatisticsConverter(org.apache.flink.table.planner.utils.CatalogTableStatisticsConverter) RexNodeExtractor(org.apache.flink.table.planner.plan.utils.RexNodeExtractor) Expression(org.apache.flink.table.expressions.Expression) Filter(org.apache.calcite.rel.core.Filter) ObjectPath(org.apache.flink.table.catalog.ObjectPath) Function(java.util.function.Function) ArrayList(java.util.ArrayList) CatalogTableStatistics(org.apache.flink.table.catalog.stats.CatalogTableStatistics) Catalog(org.apache.flink.table.catalog.Catalog) PartitionPruner(org.apache.flink.table.planner.plan.utils.PartitionPruner) ResolvedCatalogTable(org.apache.flink.table.catalog.ResolvedCatalogTable) RelDataType(org.apache.calcite.rel.type.RelDataType) JavaConversions(scala.collection.JavaConversions) TableNotExistException(org.apache.flink.table.catalog.exceptions.TableNotExistException) RexBuilder(org.apache.calcite.rex.RexBuilder) TableException(org.apache.flink.table.api.TableException) Option(scala.Option) FlinkRelOptUtil(org.apache.flink.table.planner.plan.utils.FlinkRelOptUtil) TableSourceTable(org.apache.flink.table.planner.plan.schema.TableSourceTable) RelOptRuleCall(org.apache.calcite.plan.RelOptRuleCall) CatalogPartitionSpec(org.apache.flink.table.catalog.CatalogPartitionSpec) RelOptRule(org.apache.calcite.plan.RelOptRule) FlinkStatistic(org.apache.flink.table.planner.plan.stats.FlinkStatistic) RexShuttle(org.apache.calcite.rex.RexShuttle) CatalogException(org.apache.flink.table.catalog.exceptions.CatalogException) ObjectPath(org.apache.flink.table.catalog.ObjectPath) LogicalType(org.apache.flink.table.types.logical.LogicalType) RelDataType(org.apache.calcite.rel.type.RelDataType) FlinkStatistic(org.apache.flink.table.planner.plan.stats.FlinkStatistic) RexBuilder(org.apache.calcite.rex.RexBuilder) List(java.util.List) ArrayList(java.util.ArrayList) TableSourceTable(org.apache.flink.table.planner.plan.schema.TableSourceTable) ObjectIdentifier(org.apache.flink.table.catalog.ObjectIdentifier) TableException(org.apache.flink.table.api.TableException) RelBuilder(org.apache.calcite.tools.RelBuilder) FlinkContext(org.apache.flink.table.planner.calcite.FlinkContext) TableStats(org.apache.flink.table.plan.stats.TableStats) LogicalTableScan(org.apache.calcite.rel.logical.LogicalTableScan) Catalog(org.apache.flink.table.catalog.Catalog) PartitionPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.PartitionPushDownSpec) Filter(org.apache.calcite.rel.core.Filter) Map(java.util.Map) Seq(scala.collection.Seq) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) RexNode(org.apache.calcite.rex.RexNode)

Example 25 with DynamicTableSource

use of org.apache.flink.table.connector.source.DynamicTableSource in project flink by apache.

the class PushPartitionIntoTableSourceScanRule method matches.

@Override
public boolean matches(RelOptRuleCall call) {
    Filter filter = call.rel(0);
    if (filter.getCondition() == null) {
        return false;
    }
    TableSourceTable tableSourceTable = call.rel(1).getTable().unwrap(TableSourceTable.class);
    if (tableSourceTable == null) {
        return false;
    }
    DynamicTableSource dynamicTableSource = tableSourceTable.tableSource();
    if (!(dynamicTableSource instanceof SupportsPartitionPushDown)) {
        return false;
    }
    CatalogTable catalogTable = tableSourceTable.contextResolvedTable().getTable();
    if (!catalogTable.isPartitioned() || catalogTable.getPartitionKeys().isEmpty()) {
        return false;
    }
    return Arrays.stream(tableSourceTable.abilitySpecs()).noneMatch(spec -> spec instanceof PartitionPushDownSpec);
}
Also used : PartitionPushDownSpec(org.apache.flink.table.planner.plan.abilities.source.PartitionPushDownSpec) Filter(org.apache.calcite.rel.core.Filter) TableSourceTable(org.apache.flink.table.planner.plan.schema.TableSourceTable) CatalogTable(org.apache.flink.table.catalog.CatalogTable) ResolvedCatalogTable(org.apache.flink.table.catalog.ResolvedCatalogTable) DynamicTableSource(org.apache.flink.table.connector.source.DynamicTableSource) SupportsPartitionPushDown(org.apache.flink.table.connector.source.abilities.SupportsPartitionPushDown)

Aggregations

DynamicTableSource (org.apache.flink.table.connector.source.DynamicTableSource)55 Test (org.junit.Test)24 DynamicTableSink (org.apache.flink.table.connector.sink.DynamicTableSink)12 TestDynamicTableFactory (org.apache.flink.table.factories.TestDynamicTableFactory)12 Test (org.junit.jupiter.api.Test)10 RowData (org.apache.flink.table.data.RowData)9 DecodingFormatMock (org.apache.flink.table.factories.TestFormatFactory.DecodingFormatMock)8 TableSourceTable (org.apache.flink.table.planner.plan.schema.TableSourceTable)8 ResolvedSchema (org.apache.flink.table.catalog.ResolvedSchema)7 HashMap (java.util.HashMap)5 Configuration (org.apache.flink.configuration.Configuration)5 ScanTableSource (org.apache.flink.table.connector.source.ScanTableSource)5 ParameterizedTest (org.junit.jupiter.params.ParameterizedTest)5 ArrayList (java.util.ArrayList)4 JdbcConnectorOptions (org.apache.flink.connector.jdbc.internal.options.JdbcConnectorOptions)4 JdbcLookupOptions (org.apache.flink.connector.jdbc.internal.options.JdbcLookupOptions)4 CatalogTable (org.apache.flink.table.catalog.CatalogTable)4 SourceAbilitySpec (org.apache.flink.table.planner.plan.abilities.source.SourceAbilitySpec)4 List (java.util.List)3 LogicalTableScan (org.apache.calcite.rel.logical.LogicalTableScan)3