snowflake-sql：min vs first_value windows函数的情况

Posted 2023-03-17

【中文标题】snowflake-sql：min vs first_value windows函数的情况

【英文标题】：snowflake-sql: case when min vs first_value windows function

【发布时间】：2021-12-03 02:23:54

【问题描述】：

我希望有人可以提供帮助。假设我有这张桌子

sto_id	pro_id	o_date	d_date
38	111	2021-04-01
30	111	2021-04-10	2021-04-10
30	222	2021-04-15
30	222	2021-04-18	2021-04-19
30	333	2021-04-20

而我想看到的结果是

first_sto_id	pro_id	min_o_date	first_d_date
38	111	2021-04-01
30	222	2021-04-15	2021-04-19
30	333	2021-04-20

因此，如果 sto_id 为 30 并且 d_date 为空，我想要查看此 pro_id 的下一个 d_date（如果可用）。如果 sto_id 不是 30 那么 d_date 即使它是空的。

我试过这个：

select 
first_value(sto_id) OVER (PARTITION BY pro_id ORDER BY o_date ASC) as first_sto_id,
pro_id,
min(o_date) OVER (PARTITION BY pro_id ORDER BY o_date ASC) as min_o_date,
case when sto_id='30' then min(d_date) OVER (PARTITION BY pro_id ORDER BY o_date ASC)
else first_value(d_date) OVER (PARTITION BY pro_id ORDER BY o_date ASC) end as first_d_date
from tab1

但我没有得到我需要的结果。这里还有fiddle

有什么建议吗？？ 提前非常感谢。

【参考方案1】：

使用数组：

-- PostgreSQL
SELECT
 (ARRAY_AGG(sto_id ORDER BY o_date))[1] AS first_sto_id
,pro_id
,MIN(o_date) AS min_o_date
,CASE WHEN (ARRAY_AGG(sto_id ORDER BY o_date))[1] != '30'
           THEN (ARRAY_AGG(d_date ORDER BY o_date))[1]
      ELSE (ARRAY_REMOVE(ARRAY_AGG(d_date ORDER BY o_date),NULL))[1]
      END
FROM tab1
GROUP BY pro_id
ORDER BY pro_id;

db<>fiddle demo

雪花版：

SELECT
 (ARRAY_AGG(sto_id) WITHIN GROUP(ORDER BY o_date))[0]::int AS first_sto_id
,pro_id
,MIN(o_date) AS min_o_date
,CASE WHEN (ARRAY_AGG(sto_id) WITHIN GROUP(ORDER BY o_date))[0] != '30'
           THEN (ARRAY_AGG(d_date) WITHIN GROUP (ORDER BY o_date))[0]::date
      ELSE (ARRAY_COMPACT(ARRAY_AGG(d_date) WITHIN GROUP(ORDER BY o_date)))[0]::date
      END
FROM tab1
GROUP BY pro_id
ORDER BY pro_id;

相关：Equivalent for Keep in Snowflake

【讨论】：

嗨@Lukasz，我得到了这种格式“30”的first_sto_id，d_date也像这种“2021-04-19”。这可以避免吗？

@pooq Cast is required :: int/::date