发现产品价格异常

Question

我遇到了一个 SQL 问题，这个问题让我很困惑，我想知道是否有人可以阐明执行此操作的最佳方法。

问题：鉴于 table “product_price” 表示一段时间内产品的平均价格。该范围内的所有日期都出现在未排序的“天”列中。

day	Price ($)
4	2
5	1
3	1
1	1
2	5

查找该时间段内产品价格的异常情况，更具体地说，查找价格飙升的日子（价格严格高于前一天和后一天的日子）。

第一天和最后一天永远不会是峰值。

预期结果：输出应按“天”升序排列

day
2
4

查询：我已经厌倦了这个查询

SELECT day FROM (
    SELECT
        day
        ,prev_price - price AS pre_diff
        ,price - next_price AS next_diff
    FROM (
        SELECT 
            day
            ,price
            ,LEAD(price) OVER(price) AS prev_price
            ,LAG(price) OVER(price) AS next_price
        FROM product_price
        ORDER BY day ASC
    ) a
) b
WHERE pre_diff > 1 and next_diff > 1
ORDER by day;

更新： 我想我已经解决了。感谢@hsnsd 的提示。如果有更好的方法来编写此查询，我们仍将不胜感激。

SELECT day FROM (
    SELECT
        day
        ,price
        ,prev_price
        ,next_price
        ,price - prev_price AS pre_diff
        ,next_price - price AS next_diff
    FROM (
        SELECT 
            day
            ,price
            ,COALESCE(LAG(price) OVER(ORDER BY day ASC),0) AS prev_price
            ,COALESCE(LEAD(price) OVER(ORDER BY day ASC),0) AS next_price
        FROM prices
    ) a
) b
WHERE (day NOT IN (SELECT MIN(day) FROM  prices) AND
day NOT IN (SELECT MAX(day) FROM  prices))
AND (pre_diff > 0 AND  next_diff < 0)
ORDER BY "day";

Answer 1

看起来下面应该会给你想要的结果：

with s as (
  select *, 
    Lag(Price, 1, Price) over(order by day) p,
    Lead(Price, 1, Price) over(order by day) n
  from prices
)
select day
from s
where p < Price and n < Price;

发现产品价格异常

Finding anomalies in the product prices

sql

postgresql