Question

我正在寻找编写此查询的PostgreSQL版本的优雅方式：

SELECT count(distinct a,b) / count(distinct c,d)
FROM t1
GROUP BY e;

它可以通过CONCAT功能完成，但在这种情况下我会获得巨大的性能下降。

每个计数不同都可以这样写：

SELECT count(*), e FROM
(SELECT a, b, e FROM t1 GROUP BY a, b, e) as tmp
GROUP BY e;

AND

SELECT count(*), e FROM
(SELECT c, d, e FROM t1 GROUP BY c, d, e) as tmp
GROUP BY e;

但我无法找到如何将它们组合成一个查询以获得最终结果

这是用于测试语法的create table：

CREATE TABLE t1 (
a int,
b int,
c int, 
d int, 
e int 
);

Answer 1

不确定优雅，但看起来可以这样做：

SELECT e, c1/c2 FROM

(SELECT count(*) as c1, e FROM
(SELECT a, b, e FROM t1 GROUP BY a, b, e) as sub1
GROUP BY e) as tmp1

INNER JOIN 

(SELECT count(*) as c2, e FROM
(SELECT c, d, e FROM t1 GROUP BY c, d, e) as sub2
GROUP BY e) as tmp2

USING (e)

PostgreSQL与其他字段

1 个答案: