dawidwys commented on a change in pull request #14329:
URL: https://github.com/apache/flink/pull/14329#discussion_r538255915
##########
File path:
flink-table/flink-table-common/src/main/java/org/apache/flink/table/types/extraction/DataTypeExtractor.java
##########
@@ -377,6 +397,11 @@ private void checkForCommonErrors(Type type) {
"Usually, this indicates that class
information is missing or got lost. " +
"Please specify a more concrete class
or treat it as a RAW type.",
Object.class.getName());
+ } else if (INTERNAL_DATA_STRUCTURES.contains(clazz)) {
+ throw extractionError(
+ "Cannot extract a data type from a pure
internal '%s' class. " +
Review comment:
very nit: Isn't it because those are parameterized types rather than
"pure internal"?
##########
File path:
flink-table/flink-table-planner-blink/src/test/scala/org/apache/flink/table/planner/plan/stream/sql/join/LookupJoinTest.scala
##########
@@ -211,24 +212,26 @@ class LookupJoinTest(legacyTableSource: Boolean) extends
TableTestBase with Seri
expectExceptionThrown(
"SELECT * FROM T JOIN LookupTable7 " +
"FOR SYSTEM_TIME AS OF T.proctime AS D ON T.a = D.id AND T.b = D.name
AND T.ts = D.ts",
- "Expected: eval(java.util.concurrent.CompletableFuture, " +
- "java.lang.Integer, org.apache.flink.table.data.StringData, " +
- "org.apache.flink.table.data.TimestampData) \n" +
- "Actual: eval(java.lang.Integer,
org.apache.flink.table.data.StringData, " +
- "java.time.LocalDateTime)",
- classOf[TableException]
+ "Could not find an implementation method 'eval' in class " +
+
"'org.apache.flink.table.planner.plan.utils.InvalidAsyncTableFunctionEvalSignature1'
" +
+ "for function 'default_catalog.default_database.LookupTable7' that
matches the " +
+ "following signature:\n" +
+ "void eval(java.util.concurrent.CompletableFuture,
java.lang.Integer, " +
+ "org.apache.flink.table.data.StringData,
org.apache.flink.table.data.TimestampData)",
Review comment:
Do we always fail with the internal data types now? Could we fail with
the default classes?
##########
File path:
flink-table/flink-table-planner-blink/src/main/scala/org/apache/flink/table/planner/codegen/LookupJoinCodeGenerator.scala
##########
@@ -116,105 +105,185 @@ object LookupJoinCodeGenerator {
def generateAsyncLookupFunction(
config: TableConfig,
typeFactory: FlinkTypeFactory,
+ dataTypeFactory: DataTypeFactory,
inputType: LogicalType,
+ tableSourceType: LogicalType,
returnType: LogicalType,
- tableReturnTypeInfo: TypeInformation[_],
- lookupKeyInOrder: Array[Int],
- allLookupFields: Map[Int, LookupKey],
- asyncLookupFunction: AsyncTableFunction[_])
- : GeneratedFunction[AsyncFunction[RowData, AnyRef]] = {
+ lookupKeys: Map[Int, LookupKey],
+ lookupKeyOrder: Array[Int],
+ asyncLookupFunction: AsyncTableFunction[_],
+ functionName: String)
+ : (GeneratedFunction[AsyncFunction[RowData, AnyRef]], DataType) = {
- val ctx = CodeGeneratorContext(config)
- val (prepareCode, parameters, nullInParameters) = prepareParameters(
- ctx,
+ generateLookupFunction(
+ classOf[AsyncFunction[RowData, AnyRef]],
+ config,
typeFactory,
+ dataTypeFactory,
inputType,
- lookupKeyInOrder,
- allLookupFields,
- tableReturnTypeInfo.isInstanceOf[RowTypeInfo],
- fieldCopy = true) // always copy input field because of async buffer
-
- val lookupFunctionTerm = ctx.addReusableFunction(asyncLookupFunction)
- val DELEGATE = className[DelegatingResultFuture[_]]
+ tableSourceType,
+ returnType,
+ lookupKeys,
+ lookupKeyOrder,
+ classOf[AsyncTableFunction[_]],
+ asyncLookupFunction,
+ functionName,
+ fieldCopy = true, // always copy input field because of async buffer
+ _.code)
+ }
- // TODO: filter all records when there is any nulls on the join key,
because
- // "IS NOT DISTINCT FROM" is not supported yet.
- val body =
- s"""
- |$prepareCode
- |if ($nullInParameters) {
- |
$DEFAULT_COLLECTOR_TERM.complete(java.util.Collections.emptyList());
- | return;
- |} else {
- | $DELEGATE delegates = new $DELEGATE($DEFAULT_COLLECTOR_TERM);
- | $lookupFunctionTerm.eval(delegates.getCompletableFuture(),
$parameters);
- |}
- """.stripMargin
+ private def generateLookupFunction[F <: Function](
+ generatedClass: Class[F],
+ config: TableConfig,
+ typeFactory: FlinkTypeFactory,
Review comment:
unused parameter
##########
File path:
flink-table/flink-table-planner-blink/src/main/scala/org/apache/flink/table/planner/codegen/calls/BridgingFunctionGenUtil.scala
##########
@@ -0,0 +1,419 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.flink.table.planner.codegen.calls
+
+import org.apache.flink.table.api.DataTypes
+import org.apache.flink.table.data.GenericRowData
+import
org.apache.flink.table.functions.UserDefinedFunctionHelper.{ASYNC_TABLE_EVAL,
SCALAR_EVAL, TABLE_EVAL, validateClassForRuntime}
+import org.apache.flink.table.functions._
+import org.apache.flink.table.planner.codegen.CodeGenUtils._
+import org.apache.flink.table.planner.codegen.GeneratedExpression.{NEVER_NULL,
NO_CODE}
+import org.apache.flink.table.planner.codegen._
+import org.apache.flink.table.planner.utils.JavaScalaConversionUtil.toScala
+import org.apache.flink.table.runtime.collector.WrappingCollector
+import
org.apache.flink.table.runtime.operators.join.lookup.DelegatingResultFuture
+import org.apache.flink.table.types.DataType
+import
org.apache.flink.table.types.extraction.ExtractionUtils.primitiveToWrapper
+import org.apache.flink.table.types.inference.{CallContext, TypeInference,
TypeInferenceUtil}
+import
org.apache.flink.table.types.logical.utils.LogicalTypeCasts.supportsAvoidingCast
+import org.apache.flink.table.types.logical.utils.LogicalTypeChecks.{hasRoot,
isCompositeType}
+import org.apache.flink.table.types.logical.{LogicalType, LogicalTypeRoot,
RowType}
+import
org.apache.flink.table.types.utils.DataTypeUtils.{validateInputDataType,
validateOutputDataType}
+import org.apache.flink.util.Preconditions
+
+import java.util.concurrent.CompletableFuture
+
+/**
+ * Helps in generating a call to a user-defined [[ScalarFunction]],
[[TableFunction]],
+ * or [[AsyncTableFunction]].
+ *
+ * Table functions are a special case because they are using a collector.
Thus, the result of the
+ * generation will be a reference to a [[WrappingCollector]]. Furthermore,
atomic types are wrapped
+ * into a row by the collector.
+ *
+ * Async table functions don't support atomic types.
+ */
+object BridgingFunctionGenUtil {
+
+ def generateFunctionAwareCall(
+ ctx: CodeGeneratorContext,
+ operands: Seq[GeneratedExpression],
+ returnType: LogicalType,
+ inference: TypeInference,
+ callContext: CallContext,
+ udf: UserDefinedFunction,
+ functionName: String,
+ skipIfArgsNull: Boolean)
+ : GeneratedExpression = {
+
+ val (call, _) = generateFunctionAwareCallWithDataType(
+ ctx,
+ operands,
+ returnType,
+ inference,
+ callContext,
+ udf,
+ functionName,
+ skipIfArgsNull)
+
+ call
+ }
+
+ def generateFunctionAwareCallWithDataType(
+ ctx: CodeGeneratorContext,
+ operands: Seq[GeneratedExpression],
+ returnType: LogicalType,
+ inference: TypeInference,
+ callContext: CallContext,
+ udf: UserDefinedFunction,
+ functionName: String,
+ skipIfArgsNull: Boolean)
+ : (GeneratedExpression, DataType) = {
+
+ // enrich argument types with conversion class
+ val adaptedCallContext = TypeInferenceUtil.adaptArguments(
+ inference,
+ callContext,
+ null)
+ val enrichedArgumentDataTypes =
toScala(adaptedCallContext.getArgumentDataTypes)
+ verifyArgumentTypes(operands.map(_.resultType), enrichedArgumentDataTypes)
+
+ // enrich output types with conversion class
+ val enrichedOutputDataType = TypeInferenceUtil.inferOutputType(
+ adaptedCallContext,
+ inference.getOutputTypeStrategy)
+ verifyFunctionAwareOutputType(returnType, enrichedOutputDataType, udf)
+
+ // find runtime method and generate call
+ verifyFunctionAwareImplementation(
+ enrichedArgumentDataTypes,
+ enrichedOutputDataType,
+ udf,
+ functionName)
+ val call = generateFunctionAwareCall(
+ ctx,
+ operands,
+ enrichedArgumentDataTypes,
+ enrichedOutputDataType,
+ returnType,
+ udf,
+ skipIfArgsNull)
+ (call, enrichedOutputDataType)
+ }
+
+ private def generateFunctionAwareCall(
+ ctx: CodeGeneratorContext,
+ operands: Seq[GeneratedExpression],
+ argumentDataTypes: Seq[DataType],
+ outputDataType: DataType,
+ returnType: LogicalType,
+ udf: UserDefinedFunction,
+ skipIfArgsNull: Boolean)
+ : GeneratedExpression = {
+
+ val functionTerm = ctx.addReusableFunction(udf)
+
+ // operand conversion
+ val externalOperands = prepareExternalOperands(ctx, operands,
argumentDataTypes)
+
+ if (udf.getKind == FunctionKind.TABLE) {
+ generateTableFunctionCall(
+ ctx,
+ functionTerm,
+ externalOperands,
+ outputDataType,
+ returnType,
+ skipIfArgsNull)
+ } else if (udf.getKind == FunctionKind.ASYNC_TABLE) {
+ generateAsyncTableFunctionCall(
+ ctx,
+ functionTerm,
+ externalOperands,
+ outputDataType,
+ returnType)
+ } else {
+ generateScalarFunctionCall(ctx, functionTerm, externalOperands,
outputDataType)
+ }
+ }
+
+ private def generateTableFunctionCall(
+ ctx: CodeGeneratorContext,
+ functionTerm: String,
+ externalOperands: Seq[GeneratedExpression],
+ functionOutputDataType: DataType,
+ outputType: LogicalType,
+ skipIfArgsNull: Boolean)
+ : GeneratedExpression = {
+ val resultCollectorTerm = generateResultCollector(ctx,
functionOutputDataType, outputType)
+
+ val setCollectorCode = s"""
+ |$functionTerm.setCollector($resultCollectorTerm);
+ |""".stripMargin
+ ctx.addReusableOpenStatement(setCollectorCode)
+
+ val functionCallCode = if (skipIfArgsNull) {
+ s"""
+ |${externalOperands.map(_.code).mkString("\n")}
+ |if (${externalOperands.map(_.nullTerm).mkString(" || ")}) {
+ | // skip
+ |} else {
+ | $functionTerm.eval(${externalOperands.map(_.resultTerm).mkString(",
")});
+ |}
+ |""".stripMargin
+ } else {
+ s"""
+ |${externalOperands.map(_.code).mkString("\n")}
+ |$functionTerm.eval(${externalOperands.map(_.resultTerm).mkString(",
")});
+ |""".stripMargin
+ }
+
+ // has no result
+ GeneratedExpression(
+ resultCollectorTerm,
+ NEVER_NULL,
+ functionCallCode,
+ outputType)
+ }
+
+ private def generateAsyncTableFunctionCall(
+ ctx: CodeGeneratorContext,
+ functionTerm: String,
+ externalOperands: Seq[GeneratedExpression],
+ functionOutputDataType: DataType,
+ outputType: LogicalType)
+ : GeneratedExpression = {
+
+ val DELEGATE = className[DelegatingResultFuture[_]]
+
+ val functionCallCode =
+ s"""
+ |${externalOperands.map(_.code).mkString("\n")}
+ |if (${externalOperands.map(_.nullTerm).mkString(" || ")}) {
+ | $DEFAULT_COLLECTOR_TERM.complete(java.util.Collections.emptyList());
+ |} else {
+ | $DELEGATE delegates = new $DELEGATE($DEFAULT_COLLECTOR_TERM);
+ | $functionTerm.eval(
+ | delegates.getCompletableFuture(),
+ | ${externalOperands.map(_.resultTerm).mkString(", ")});
+ |}
+ |""".stripMargin
+
+ // has no result
+ GeneratedExpression(
+ NO_CODE,
+ NEVER_NULL,
+ functionCallCode,
+ outputType)
+ }
+
+ /**
+ * Generates a collector that converts the output of a table function
(possibly as an atomic type)
+ * into an internal row type. Returns a collector term for referencing the
collector.
+ */
+ private def generateResultCollector(
+ ctx: CodeGeneratorContext,
+ outputDataType: DataType,
+ returnType: LogicalType)
+ : String = {
+ val outputType = outputDataType.getLogicalType
+
+ val collectorCtx = CodeGeneratorContext(ctx.tableConfig)
+ val externalResultTerm = newName("externalResult")
+
+ // code for wrapping atomic types
+ val collectorCode = if (!isCompositeType(outputType)) {
+ val resultGenerator = new ExprCodeGenerator(collectorCtx,
outputType.isNullable)
+ .bindInput(outputType, externalResultTerm)
+ val wrappedResult = resultGenerator.generateConverterResultExpression(
+ returnType.asInstanceOf[RowType],
+ classOf[GenericRowData])
+ s"""
+ |${wrappedResult.code}
+ |outputResult(${wrappedResult.resultTerm});
+ |""".stripMargin
+ } else {
+ s"""
+ |if ($externalResultTerm != null) {
+ | outputResult($externalResultTerm);
+ |}
+ |""".stripMargin
+ }
+
+ // collector for converting to internal types then wrapping atomic types
+ val resultCollector = CollectorCodeGenerator.generateWrappingCollector(
+ collectorCtx,
+ "TableFunctionResultConverterCollector",
+ outputType,
+ externalResultTerm,
+ // nullability is handled by the expression code generator if necessary
+ genToInternalConverter(ctx, outputDataType),
+ collectorCode)
+ val resultCollectorTerm = newName("resultConverterCollector")
+ CollectorCodeGenerator.addToContext(ctx, resultCollectorTerm,
resultCollector)
+
+ resultCollectorTerm
+ }
+
+ private def generateScalarFunctionCall(
+ ctx: CodeGeneratorContext,
+ functionTerm: String,
+ externalOperands: Seq[GeneratedExpression],
+ outputDataType: DataType)
+ : GeneratedExpression = {
+
+ // result conversion
+ val externalResultClass = outputDataType.getConversionClass
+ val externalResultTypeTerm = typeTerm(externalResultClass)
+ // Janino does not fully support the JVM spec:
+ // boolean b = (boolean) f(); where f returns Object
+ // This is not supported and we need to box manually.
+ val externalResultClassBoxed = primitiveToWrapper(externalResultClass)
+ val externalResultCasting = if (externalResultClass ==
externalResultClassBoxed) {
+ s"($externalResultTypeTerm)"
+ } else {
+ s"($externalResultTypeTerm) (${typeTerm(externalResultClassBoxed)})"
+ }
+ val externalResultTerm =
ctx.addReusableLocalVariable(externalResultTypeTerm, "externalResult")
+ val internalExpr = genToInternalConverterAll(ctx, outputDataType,
externalResultTerm)
+
+ // function call
+ internalExpr.copy(code =
+ s"""
+ |${externalOperands.map(_.code).mkString("\n")}
+ |$externalResultTerm = $externalResultCasting $functionTerm
+ | .$SCALAR_EVAL(${externalOperands.map(_.resultTerm).mkString(", ")});
+ |${internalExpr.code}
+ |""".stripMargin)
+ }
+
+ private def prepareExternalOperands(
+ ctx: CodeGeneratorContext,
+ operands: Seq[GeneratedExpression],
+ argumentDataTypes: Seq[DataType])
+ : Seq[GeneratedExpression] = {
+ operands
+ .zip(argumentDataTypes)
+ .map { case (operand, dataType) =>
+ operand.copy(resultTerm = genToExternalConverterAll(ctx, dataType,
operand))
+ }
+ }
+
+ private def verifyArgumentTypes(
+ operandTypes: Seq[LogicalType],
+ enrichedDataTypes: Seq[DataType])
+ : Unit = {
+ val enrichedTypes = enrichedDataTypes.map(_.getLogicalType)
+ operandTypes.zip(enrichedTypes).foreach { case (operandType, enrichedType)
=>
+ // check that the logical type has not changed during the enrichment
+ if (!supportsAvoidingCast(operandType, enrichedType)) {
+ throw new CodeGenException(
+ s"Mismatch of function's argument data type '$enrichedType' and
actual " +
+ s"argument type '$operandType'.")
+ }
+ }
+ // the data type class can only partially verify the conversion class,
+ // now is the time for the final check
+ enrichedDataTypes.foreach(validateOutputDataType)
+ }
+
+ private def verifyFunctionAwareOutputType(
+ returnType: LogicalType,
+ enrichedDataType: DataType,
+ udf: UserDefinedFunction)
+ : Unit = {
+ val enrichedType = enrichedDataType.getLogicalType
+ if (udf.getKind == FunctionKind.TABLE && !isCompositeType(enrichedType)) {
+ // logically table functions wrap atomic types into ROW, however, the
physical function might
+ // return an atomic type
+ Preconditions.checkState(
+ hasRoot(returnType, LogicalTypeRoot.ROW) &&
returnType.getChildren.size() == 1,
+ "Logical output type of function call should be a ROW wrapping an
atomic type.",
+ Seq(): _*)
+ val atomicOutputType =
returnType.asInstanceOf[RowType].getChildren.get(0)
+ verifyOutputType(atomicOutputType, enrichedDataType, udf)
+ } else if (udf.getKind == FunctionKind.ASYNC_TABLE &&
!isCompositeType(enrichedType)) {
+ throw new CodeGenException(
+ "Async table functions must not emit an atomic type. " +
+ "Only a composite type such as the row type are supported.")
+ }
+ else if (udf.getKind == FunctionKind.TABLE || udf.getKind ==
FunctionKind.ASYNC_TABLE) {
+ // null values are skipped therefore, the result top level row will
always be not null
+ verifyOutputType(returnType.copy(true), enrichedDataType, udf)
+ } else {
+ verifyOutputType(returnType, enrichedDataType, udf)
+ }
+ }
+
+ private def verifyOutputType(
+ returnType: LogicalType,
+ enrichedDataType: DataType,
+ udf: UserDefinedFunction)
+ : Unit = {
+ val enrichedType = enrichedDataType.getLogicalType
+ // check that the logical type has not changed during the enrichment
+ if (!supportsAvoidingCast(enrichedType, returnType)) {
+ throw new CodeGenException(
+ s"Mismatch of expected output data type '$returnType' and function's "
+
+ s"output type '$enrichedType'.")
+ }
+ // the data type class can only partially verify the conversion class,
+ // now is the time for the final check
+ validateInputDataType(enrichedDataType)
+ }
+
+ def verifyFunctionAwareImplementation(
+ argumentDataTypes: Seq[DataType],
+ outputDataType: DataType,
+ udf: UserDefinedFunction,
+ functionName: String)
+ : Unit = {
+ if (udf.getKind == FunctionKind.TABLE) {
+ verifyImplementation(TABLE_EVAL, argumentDataTypes, None, udf,
functionName)
+ } else if (udf.getKind == FunctionKind.ASYNC_TABLE) {
+ verifyImplementation(
+ ASYNC_TABLE_EVAL,
+ Seq(DataTypes.NULL.bridgedTo(classOf[CompletableFuture[_]])) ++
argumentDataTypes,
Review comment:
nit:
`DataTypes.NULL.bridgedTo(classOf[CompletableFuture[_]]) +:
argumentDataTypes`
I might be wrong, but I think this will create one less object.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
