表值函数 (TVF) 调用

适用于: 勾选“是” Databricks SQL 勾选“是” Databricks Runtime

调用一个函数,该函数返回一个关系或一组行作为 [table-reference](sql-ref

TVF 可以是:

  • SQL 用户定义的表函数。

  • range 表值函数。

  • 任何表值生成器函数,例如 explode

    适用于:勾选“是” Databricks SQL 勾选“是” Databricks Runtime 12.2 LTS 及更高版本。

语法

function_name ( [ expression [, ...] ] ) [ table_alias ]

参数

  • function_name

    表值函数。

  • expression

    生成值的一个或多个值、运算符和 SQL 函数的组合。

  • table_alias

    用于引用函数结果及其列的可选标签。

示例

-- range call with end
> SELECT * FROM range(6 + cos(3));
   0
   1
   2
   3
   4

-- range call with start and end
> SELECT * FROM range(5, 10);
   5
   6
   7
   8
   9

-- range call with numPartitions
> SELECT * FROM range(0, 10, 2, 200);
   0
   2
   4
   6
   8

-- range call with a table alias
> SELECT * FROM range(5, 8) AS test;
   5
   6
   7

-- Create a SQL UDTF and invoke it
> CREATE OR REPLACE FUNCTION table_func(a INT) RETURNS TABLE
    RETURN SELECT a * c1 AS res FROM VALUES(1), (2), (3), (4) AS T(c1)

> SELECT * FROM table_func(5);
   5
  10
  15
  20

-- Using lateral correlation
>  SELECT table_func.res FROM VALUES(10), (20) AS S(c1), LATERAL table_func(c1);
  10
  20
  20
  40
  30
  60
  40
  80

-- Scalar fucntion are not allowed in the FROM clause
> SELECT * FROM trim('hello  ');
  Error

在 Databricks SQL 和 Databricks Runtime 12.2 LTS 及更高版本上:

> SELECT * FROM explode(array(10, 20));
  10
  20

> SELECT * FROM inline(array(struct(1, 'a'), struct(2, 'b')));
 col1 col2
 ---- ----
    1    a
    2    b

> SELECT * FROM posexplode(array(10,20));
 pos col
 --- ---
   0  10
   1  20

> SELECT * FROM stack(2, 1, 2, 3);
 col0 col1
 ---- ----
    1    2
    3 null

> SELECT * FROM json_tuple('{"a":1, "b":2}', 'a', 'b');
  c0  c1
 --- ---
   1   2

> SELECT * FROM parse_url('http://spark.apache.org/path?query=1', 'HOST');
  spark.apache.org

> SELECT * FROM VALUES(1), (2) AS t1(c1), LATERAL explode (ARRAY(3,4)) AS t2(c2);
 c1 c2
 -- --
  1  3
  1  4
  2  3
  2  4