13: SQL client 使用 UDF demo

1: ScalarFunction： 这种类型的UDF是一对一的，输入一个值，返回一个值。

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

import org.apache.flink.table.functions.ScalarFunction;       public class Hashstring extends ScalarFunction {         public Hashstring(){};     private int factor = 12;     public Hashstring(int factor)     {         this.factor = factor;     }     public int eval(String s) {         return s.hashCode() * factor;     } }

2： TableFunction 是一对多的。下面的demo是输入一个String类型，返回Row 类型，有两列

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

import org.apache.flink.table.annotation.DataTypeHint; import org.apache.flink.table.annotation.FunctionHint; import org.apache.flink.table.functions.TableFunction; import org.apache.flink.types.Row;     @FunctionHint(output = @DataTypeHint("ROW<first STRING, name STRING>")) public  class MyFunction extends TableFunction<Row> {       public void eval(String value) {         Row row = new Row(2);         row.setField(0, "first");         row.setField(1, value.toString());         collect(row);     } }

3：

 启动SQL client:               ./bin/sql-client.sh -embedded

下面是SQL client里执行的脚本；

?

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

create temporary function hash_compute as 'Hashstring'; create temporary function my_function as 'MyFunction';   CREATE TABLE orders (     order_uid  BIGINT,     product_name String,     price      DECIMAL(32, 2),     order_time TIMESTAMP(3) ) WITH (     'connector' = 'datagen' );      select hash_compute(product_name) from orders;   select col1, col2 from orders a LEFT JOIN LATERAL TABLE(my_function(product_name)) AS T(col1, col2) ON TRUE;