具有每个分区的最小值和最大值的雪花 SQL 行

Question

我需要找到 table 个分区上求和列的最大值和最小值。

内部查询是：

select 
ss_store_sk,
d.d_year,
d.d_moy,
sum(ss_quantity) as total_sales,
rank() over (partition by ss_store_sk order by sum(ss_quantity) desc) as "rank"
from store_sales
join date_dim as d on d.d_date_sk = ss_sold_date_sk
where d.d_year != 2003 and d.d_moy != 1
group by 
ss_store_sk,
d.d_year,
d.d_moy

这将产生如下所示的 table。

SS_STORE_SK	D_YEAR	D_MOY	TOTAL_SALES	排名
182	1999	12	60836090	1
182	1998	11	60792623	2
182	2001	10	60615582	3
182	2000	9	60459371	4
18	1998	12	232323	1
18	2001	11	123244	2
18	2000	10	3422	3

我可以通过以下方式获得 TOTAL_SALES 的最大值行：

with minmax as (
inner query
)
select * from minmax where "rank" =1

但是如何为每个 SS_STORE_SK 获取 TOTAL_SALES 的最小值行？我需要的结果如下所示。但是只要能够分别获得 TOTAL_SALES 的最小排名行就足够了。

SS_STORE_SK	D_YEAR	D_MOY	TOTAL_SALES	排名
182	1999	12	60836090	1
182	2000	9	60459371	4
18	1998	12	232323	1
18	2000	10	3422	3

我正在使用 Snowflake SQL。

Answer 1

使用max() window函数：

select ss_store_sk, d.d_year, d.d_moy,
       sum(ss_quantity) as total_sales,
       rank() over (partition by ss_store_sk order by sum(ss_quantity) desc) as "rank",
       max(sum(ss_quantity)) over (partition by ss_store_sk)
from store_sales join
     date_dim d
     on d.d_date_sk = ss_sold_date_sk
where d.d_year <> 2003 and d.d_moy <> 1
group by ss_store_sk, d.d_year, d.d_moy;

当然，如果您想要最小值，可以使用 min()。

Answer 2

使用rank() window函数：一个按sum(ss_quantity)降序排列，一个按升序排列。然后只是 select 两个排名的第一行。

with minmax as
(
   select 
   ss_store_sk,
   d.d_year,
   d.d_moy,
   sum(ss_quantity) as total_sales,
   rank() over (partition by ss_store_sk order by sum(ss_quantity) desc) as "rank"
   rank() over (partition by ss_store_sk order by sum(ss_quantity) ) as "rank2"
   from store_sales
   join date_dim as d on d.d_date_sk = ss_sold_date_sk
   where d.d_year != 2003 and d.d_moy != 1
   group by 
   ss_store_sk,
   d.d_year,
   d.d_moy
)
select * from minmax where rank =1 or rank2=1

Answer 3

一种更简洁的基于窗口函数过滤行的方法：

select 
  ss_store_sk,
  d.d_year,
  d.d_moy,
  sum(ss_quantity) as total_sales
from store_sales
join date_dim as d on d.d_date_sk = ss_sold_date_sk
where d.d_year != 2003 and d.d_moy != 1
group by ss_store_sk, d.d_year, d.d_moy
qualify rank() over (partition by ss_store_sk order by total_sales desc) = 1
     or rank() over (partition by ss_store_sk order by total_sales) = 1

具有每个分区的最小值和最大值的雪花 SQL 行

Snowflake SQL rows with minimum and maximum values for each partition

sql

snowflake-cloud-data-platform

rank