从 Scala 代码调用 java 通用 Java 方法时类型不匹配

Question

I have a n-quantity of Java classes with one superclass - data model. The list of the classes is input parameter to Scala method in which I want to create resultStreams and have to create call Java generic methods from process method. Could you write how to resolve it? I tried to use [_ <: SpecificRecordBase], [SpecificRecordBase] in the method calling, but had the same result.

Error

Error:(146, 88) type mismatch;
 found   : Class[_$3] where type _$3 <: org.apache.avro.specific.SpecificRecordBase
 required: Class[org.apache.avro.specific.SpecificRecordBase]
Note: _$3 <: org.apache.avro.specific.SpecificRecordBase, but Java-defined class Class is invariant in type T.
You may wish to investigate a wildcard type such as `_ <: org.apache.avro.specific.SpecificRecordBase`. (SLS 3.2.10)
                AvroHelper.deSerializeAvroObject(record.value, cl))(TypeInformation.of(cl)))

Scala code

object GenerickRunnerStackOverFlow  {
  def process(inputClasses :  List[Class[_ <: SpecificRecordBase]],): Unit = {
    val newStream: DataStream[KafkaSourceType] = env.addSource(....)).uid(...).filter(...)

    val resultStreams = inputClasses .map(
      cl => newStream.map(record =>
                AvroHelper.deSerializeAvroObject(record.value, cl))(TypeInformation.of(cl)))

        ...
  }

    def main(args: Array[String]): Unit = {
        val topicToClasses: List[Class[_ <: SpecificRecordBase]] = List(Types.RECORD_1.getClassType, Types.RECORD_1.getClassType.getClassType)
        process(topicToClasses)

    }
}

Java method spec

public static <A extends SpecificRecord> A deSerializeAvroObject(byte[] object, Class<A> clazz){ ...}

Model

    public class Record1 extends SpecificRecordBase {}
    public class Record2 extends SpecificRecordBase {}
    ...
    public enum Types {
      RECORD_1(Record1.class),
      RECORD_2(Record2.class);
      ....

      private Class<? extends SpecificRecordBase> clazz;
      public Class<? extends SpecificRecordBase> getClassType() {return this.clazz;}
}

Also I have the same message error with Scala addSink method:

def addSink(sinkFunction : org.apache.flink.streaming.api.functions.sink.SinkFunction[T]) : org.apache.flink.streaming.api.datastream.DataStreamSink[T] = { /* compiled code */ }

I write the wrapper method:

def addSinkWithSpecificRecordBase[A <: SpecificRecordBase](
    stream: DataStream[A],
    sink: BucketingSink[A]): DataStreamSink[A] = stream.addSink(sink)

In result of execution:

val result = topicToSinkStream.foreach { el =>
  val stream: DataStream[_ <: SpecificRecordBase] = el._2._1
  val sink: BucketingSink[_ <: SpecificRecordBase] = el._2._2
  addSinkWithSpecificRecordBase(stream, sink)
}

There was an error:

Error:(209, 37) type mismatch;
 found   : org.apache.flink.streaming.api.scala.DataStream[_$9] where type _$9 <: org.apache.avro.specific.SpecificRecordBase
 required: org.apache.flink.streaming.api.scala.DataStream[org.apache.avro.specific.SpecificRecordBase]
Note: _$9 <: org.apache.avro.specific.SpecificRecordBase, but class DataStream is invariant in type T.
You may wish to define T as +T instead. (SLS 4.5)
      addSinkWithSpecificRecordBase(stream, sink)

Where topicToSinkStream is:

Map[String, (DataStream[_ <: SpecificRecordBase], BucketingSink[_ <: SpecificRecordBase])]

I also tried to eliminate the SpecificRecordBase in method gemeric description and Add + and - to method parameters description. But there is no result.

Answer 1

The issue is that the type of AvroHelper.deSerializeAvroObject(record.value, cl) is SpecificRecordBase ( _ <: SpecificRecordBase is only allowed in type parameters , not here). The fix is to extract a helper function:

def processClass[A <: SpecificRecordBase](cl: Class[A], newStream: DataStream[KafkaSourceType]) = 
  newStream.map(record => AvroHelper.deSerializeAvroObject(record.value, cl))(TypeInformation.of(cl)))

(if you define it locally you can also use newStream without making it an argument) and then

val resultStreams = inputClasses.map(cl => processClass(cl, newStream))