在Postgres中，percentile_disc是否重用同一查询中其他percentile_disc的计算？

Question

假设您有一些正在查询百分位数的表，两列使用相同的分组：

SELECT percentile_disc(0.25) WITHIN GROUP (ORDER BY my_col) as p25,
       percentile_disc(0.75) WITHIN GROUP (ORDER BY my_col) as p75 from my_table

Postgres 在计算第二列时是否重用为第一列计算的分组？

Answer 1

确实如此。您可以在

explain analyse verbose

中看到：无论添加多少个，都只会循环一次。这并不意味着您可以免费添加它们 - 执行时间将会增加。 _{db<>fiddle 的演示}

explain analyse verbose
SELECT percentile_disc(0.25) WITHIN GROUP (ORDER BY my_col) as p25
      ,percentile_disc(0.75) WITHIN GROUP (ORDER BY my_col) as p75
      ,percentile_disc(0.3) WITHIN GROUP (ORDER BY my_col) as p
    ,percentile_disc(0.4) WITHIN GROUP (ORDER BY my_col)
    ,percentile_disc(0.5) WITHIN GROUP (ORDER BY my_col)
    ,percentile_disc(0.6) WITHIN GROUP (ORDER BY my_col)
    ,percentile_disc(0.7) WITHIN GROUP (ORDER BY my_col)
from my_table;

聚合（成本=8464.74..8464.75行=1宽度=56）（实际时间=343.493..343.495行=1循环=1）
输出：percentile_disc('0.25'::双精度)在组内(ORDER BY my_col)、percentile_disc('0.75'::双精度)在组内(ORDER BY my_col)、percentile_disc('0.3'::双精度)在组内组（ORDER BY my_col），percentile_disc（'0.4'::双精度）组内（ORDER BY my_col），percentile_disc（'0.5'::双精度）组内（ORDER BY my_col），percentile_disc（'0.6'::双精度）组内（ORDER BY my_col），percentile_disc（'0.7'::双精度）组内（ORDER BY my_col）
-> public.my_table 上的顺序扫描（成本=0.00..7214.38行=500138宽度=8）（实际时间=0.016..39.669行=500000循环=1）
输出：my_col
规划时间：0.122毫秒
执行时间：344.358 ms