使用Table AggregateFunction和ResultTypeQueryable时的ValidationException

Question

我正在使用配置为使用flink-table jar的本地Flink 1.6集群（意味着我的程序的jar不包括flink-table）。使用以下代码

import org.apache.flink.api.common.typeinfo.TypeHint;
import org.apache.flink.api.common.typeinfo.TypeInformation;
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.flink.api.java.operators.DataSource;
import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.api.java.typeutils.ResultTypeQueryable;
import org.apache.flink.table.api.Table;
import org.apache.flink.table.api.TableEnvironment;
import org.apache.flink.table.api.java.BatchTableEnvironment;
import org.apache.flink.table.functions.AggregateFunction;
import org.apache.flink.types.Row;

import java.util.ArrayList;
import java.util.List;

public class JMain {
    public static void main(String[] args) throws Exception {
        ExecutionEnvironment execEnv = ExecutionEnvironment.getExecutionEnvironment();
        BatchTableEnvironment tableEnv = TableEnvironment.getTableEnvironment(execEnv);

        tableEnv.registerFunction("enlist", new Enlister());

        DataSource<Tuple2<String, String>> source = execEnv.fromElements(
                new Tuple2<>("a", "1"),
                new Tuple2<>("a", "2"),
                new Tuple2<>("b", "3")
        );

        Table table = tableEnv.fromDataSet(source, "a, b")
                .groupBy("a")
                .select("enlist(a, b)");

        tableEnv.toDataSet(table, Row.class)
                .print();
    }

    public static class Enlister
            extends AggregateFunction<List<String>, ArrayList<String>>
            implements ResultTypeQueryable<List<String>>
    {
        @Override
        public ArrayList<String> createAccumulator() {
            return new ArrayList<>();
        }

        @Override
        public List<String> getValue(ArrayList<String> acc) {
            return acc;
        }

        @SuppressWarnings("unused")
        public void accumulate(ArrayList<String> acc, String a, String b) {
            acc.add(a + ":" + b);
        }

        @SuppressWarnings("unused")
        public void merge(ArrayList<String> acc, Iterable<ArrayList<String>> it) {
            for (ArrayList<String> otherAcc : it) {
                acc.addAll(otherAcc);
            }
        }

        @SuppressWarnings("unused")
        public void resetAccumulator(ArrayList<String> acc) {
            acc.clear();
        }

        @Override
        public TypeInformation<List<String>> getProducedType() {
            return TypeInformation.of(new TypeHint<List<String>>(){});
        }
    }
}

我得到这个奇怪的例外：

org.apache.flink.table.api.ValidationException: Expression Enlister(List('a, 'b)) failed on input check: Given parameters do not match any signature. 
Actual: (java.lang.String, java.lang.String) 
Expected: (java.lang.String, java.lang.String)

但是，如果我没有实现ResultTypeQueryable，我得到预期的输出：

Starting execution of program
[b:3]
[a:1, a:2]
Program execution finished
Job with JobID 20497bd3efe44fab0092a05a8eb7d9de has finished.
Job Runtime: 270 ms
Accumulator Results: 
- 56e0e5a9466b84ae44431c9c4b7aad71 (java.util.ArrayList) [2 elements]

我的实际用例似乎需要ResultTypeQueryable，否则我得到这个例外：

The return type of function ... could not be determined automatically,
due to type erasure. You can give type information hints by using the
returns(...) method on the result of the transformation call,
or by letting your function implement the 'ResultTypeQueryable' interface

我能解决这个问题吗？

Answer 1

在这种情况下，实施ResultTypeQueryable是不正确的。例外是误导性的。而是覆盖getResultType()和getAccumulatorType()。这背后的原因是，在为序列化器生成类型信息时，泛型通常会导致问题（由于Java的类型擦除）。

Answer 2

我尝试在一个小程序中重现问题，但我不能，它只发生在我的大项目中。不幸的是，重写getResultType()和getAccumulatorType()也没有帮助，在这种情况下我得到了这个例外：

java.lang.IndexOutOfBoundsException
    at org.apache.flink.api.java.typeutils.TupleTypeInfoBase.getTypeAt(TupleTypeInfoBase.java:199)
    at org.apache.flink.api.java.typeutils.RowTypeInfo.getTypeAt(RowTypeInfo.java:179)
    at org.apache.flink.api.common.operators.Keys$ExpressionKeys.isSortKey(Keys.java:444)
    at org.apache.flink.api.java.operators.SortPartitionOperator.ensureSortableKey(SortPartitionOperator.java:150)
    at org.apache.flink.api.java.operators.SortPartitionOperator.<init>(SortPartitionOperator.java:75)
    at org.apache.flink.api.java.DataSet.sortPartition(DataSet.java:1414)

即使没有覆盖，我实际上也得到了这个例外。对我有用的唯一事情基本上是：

String[] fieldNames = new String[] {
        "result"
};
TypeInformation<?>[] types = new TypeInformation[] {
        TypeInformation.of(new TypeHint<List<String>>(){})
};

tableEnv.toDataSet(table, Types.ROW(fieldNames, types))...