sql - Snowflake中的语句绑定(bind)/执行顺序

Question

我喜欢 Snowflake 的原因之一是，它对于子句何时可用于其他子句没有那么严格。例如以下内容:

WITH tbl (name, age) as (
    SELECT * FROM values ('david',10), ('tom',20)
)
select
    name,
    age,
    year(current_timestamp())-age as birthyear
from
    tbl
where
    birthyear > 2010

我可以在 WHERE 子句中使用 birthyear。这将与 SQL Server 之类的东西结合起来，其中绑定(bind)要严格得多，例如 here 。是否有关于 Snowflake 中绑定(bind)顺序的任何文档，或者通常是“松散”的，解析器会为您解决问题？

或者，与MySQL5.7中相同:

select
    name,
    age,
    year(now())-age as birthyear
from
    (select 'tom' name, 20 age union select 'david', 10) tbl
where birthyear > 2010

Unknown column 'birthyear' in 'where clause'

Answer 1

免责声明:这是我理解此功能的方式，可能不完整

---

由于出生年份未知，像下面这样的构造将会失败。

select
    name,
    age,
    year(current_timestamp())-age as birthyear
from  tbl
where birthyear > 2010;

通常我们使用子查询或简单地复制表达式:

select *
from ( select
          name,
          age,
          year(current_timestamp())-age as birthyear
      from  tbl) s
where birthyear > 2010;

select
    name,
    age,
    year(current_timestamp())-age as birthyear
from  tbl
where year(current_timestamp())-age > 2010;

---

还有另一种方法可以使用 LATERAL JOIN/CROSS APPLY 来实现它:

select
    name,
    age,
    s.birthyear
from  tbl
,LATERAL(SELECT year(current_timestamp())-age as birthyear) s
where s.birthyear > 2010;

我们可以更进一步，在同一级别使用两次别名:

select
    name,
    age,
    year(current_timestamp())-age as birthyear,
    CASE birthyear > 2000 THEN 'XXI'ELSE 'XX' END AS century
from  tbl
where birthyear > 2010;

这相当于链式横向连接:

select
        name,
        age,
        s.birthyear,
        s2.century
 from  tbl
 ,LATERAL (SELECT year(current_timestamp())-age as birthyear) s
 ,LATERAL (SELECT CASE WHEN s.birthyear > 2000 THEN 'XXI'ELSE 'XX' END AS century) s2
 where s.birthyear > 2010;

<强> db<>fiddle demo

Snowflake 的版本是一种“LATERAL JOIN 的内联版本”，非常简洁。

---

相关:PostgreSQL: using a calculated column in the same query

值得一提的是:此语法不允许嵌套窗口函数 Using nested window function in Snowflake或者在 WHERE 中使用它们:

SELECT *, ROW_NUMBER() OVER(ORDER BY col1) AS rn
FROM t
WHERE rn > 10;

=>
SELECT *, ROW_NUMBER() OVER(ORDER BY col1) AS rn
FROM t
QUALIFY rn > 10;