如何在 Postgres 中使用带有多个分位数的 percentile_conts

Posted 2023-02-16

【中文标题】如何在 Postgres 中使用带有多个分位数的 percentile_conts

【英文标题】：How to use percentile_conts with multiple quantiles in Postgres

【发布时间】：2021-06-14 04:02:37

【问题描述】：

我目前有一个这样的查询：

select AVG(t2 - t1) as delay,
       percentile_cont(0.25) within group (order by (t2 - t1)) as q25,
       percentile_cont(0.5) within group (order by (t2 - t1)) as median,
       percentile_cont(0.75) within group (order by (t2 - t1)) as q75,
       p.bool1,
       p.cat1
from people p
group by p.bool1, p.cat1
order by p.cat1,p.bool1

但是，我在 postgres 函数聚合页面上阅读： https://www.postgresql.org/docs/9.4/functions-aggregate.html

我应该能够指定多个分位数：

percentile_cont(fractions) WITHIN GROUP (ORDER BY sort_expression)  double precision[]  double precision or interval    array of sort expression's type     multiple continuous percentile: returns an array of results matching the shape of the fractions parameter, with each non-null element replaced by the value corresponding to that percentile

我想使用它，所以我不会为每个分位数重新计算 t2 - t1。获得多个分位数的正确语法是什么？我需要子查询吗？

【参考方案1】：

我想使用它，所以我不会为每个分位数重新计算 t2 - t1

横向连接在这种情况下会有所帮助：

select AVG(t2 - t1) as delay,
       percentile_cont(0.25) within group (order by s.col) as q25,
       percentile_cont(0.5) within group (order by s.col) as median,
       percentile_cont(0.75) within group (order by s.col) as q75,
       p.bool1,
       p.cat1
from people p
,LATERAL(SELECT t2 - t1 AS col) s
group by p.bool1, p.cat1
order by p.cat1,p.bool1;

相关：PostgreSQL: using a calculated column in the same query

---

数组定义为：ARRAY[0.25, 0.5, 0.75] 或'0.25, 0.5, 0.75'::double precision[]

select AVG(t2 - t1) as delay,
   -- 1
   percentile_cont(ARRAY[0.25, 0.5, 0.75]) within group (order by (t2 - t1)) as q25,
   -- 2
   percentile_cont('0.25, 0.5, 0.75'::double precision[]) 
   within group (order by (t2 - t1)) as q
       p.bool1,
       p.cat1
from people p
group by p.bool1, p.cat1
order by p.cat1,p.bool1;

db<>fiddle demo

---

有没有一种简单的方法可以内联指定每个生成的百分位数字段的名称，就像它们在 q25、q50、q75 等中一样

WITH cte AS (
    select AVG(t2 - t1) as delay,
       percentile_cont(ARRAY[0.25, 0.5, 0.75]) within group (order by (t2 - t1)) as q,
           p.bool1,
           p.cat1
    from people p
    group by p.bool1, p.cat1
)
select cte.*, q[1] AS q25, q[2] AS q50, q[3] AS q75
from cte
order by cat1,bool1;

db<>fiddle demo2

【讨论】：

谢谢！数组声明是我需要的关键洞察力。我不明白为什么它不能使用可变参数的双打列表！

是否有一种简单的方法可以内联指定每个生成的百分位数字段的名称，就像它们与 q25、q50、q75 等一样？

@Mittenchops 是的，您可以访问结果数组的每个元素并添加别名