如何在 flink sql 中将多集数据类型转换为字符串？

Question

when i was doing such query in flink sql:

    SELECT COLLECT(col1) OVER (
        PARTITION BY col2
        ORDER BY col3
        ROWS BETWEEN 1 PRECEDING AND CURRENT ROW
    ) AS col4
    FROM table 

how can i cast the col4 , which is a multiset datatype, to string?

i have tried cast(col4 as string) , but it didnt work. the exception is Cast function cannot convert value of type BIGINT MULTISET to type VARCHAR(2147483647)

OR how can i pass the multiset data to a java udf and then trasform to a string? how to write such udf?

Answer 1

Currently, casting of multisets is limited. The community is currently working on improving this .

Until then, I would recommend to use a scalar function. UDFs can accept all types. Since the automatic reflection logic reserves the Map class for the MAP type. You have to add a type hint.

public class MultisetToString extends ScalarFunction {

  public String eval(@DataTypeHint("MULTISET<STRING>") Map<String, Integer> multiset) {
    return multiset.toString();
  }
}

Answer 2

还有另一个悬而未决的问题上积极努力，这与打印的支持，但也铸造所有的结构类型为STRING做