sql—聚合数,有些在范围内

gc0ot86w  于 2021-07-24  发布在  Java
关注(0)|答案(2)|浏览(370)

我有一张table tab1 在我的数字列中使用以下值 c1 :

1
2
3
4
5
6
7
8
9
10
13
18

我想回来

1-5
6-10
13
18

以下查询在mysql中执行成功,但在postgresql中执行不成功:

select distinct
case
when c1 between 1 and 5 then '1-5'
when c1 between 6 and 9 then '6-10'
else c1
end as 'c1'
from tab1;
z9ju0rcb

z9ju0rcb1#

相同的查询在修复语法错误和正确的强制转换后在postgres中工作:

SELECT DISTINCT
       CASE WHEN c1 BETWEEN 1 AND 5 THEN '1-5'
            WHEN c1 BETWEEN 6 AND 9 THEN '6-10'
            ELSE c1::text END AS c1 
FROM   tab1;

结果必须是字符串类型- text 是postgres的首选类型。但按字符串表示法排序对数字没有好处。因此,我建议采用这种方法进行适当的分类:

SELECT CASE WHEN c1 BETWEEN 1 AND 5 THEN '1-5'
            WHEN c1 BETWEEN 6 AND 9 THEN '6-10'
            ELSE c1::text END AS new_c1   -- avoid duplicate of source name
FROM   tab1 t
GROUP  BY new_c1
ORDER  BY min(t.c1);

db<>在这里摆弄

gpfsuwkq

gpfsuwkq2#

如果我理解正确,您将使用两个聚合级别。第一个得到范围:

select min(c1), max(c1)
from (select t.*, row_number() over (order by c1) as seqnum
      from t
     ) t
group by c1 - seqnum;

然后,第二种方法重新聚合:

select string_agg(range, ',' order by min_c1)
from (select (case when min(c1) <> max(c1) then min(c1) || '-' || max(c1) else min(c1)::text end) as range, min(c1) as min_c1
      from (select t.*, row_number() over (order by c1) as seqnum
            from t
           ) t
      group by c1 - seqnum
     ) x

这里有一个db<>小提琴说明了这个逻辑的工作原理。

相关问题