无法在没有超时的情况下聚合包含相关子查询的查询结果 - mysql 5.7

Question

下面我的初始查询运行速度相当快，大约需要 9 秒，并返回大约 25 万条记录。我想按 Month_year 和 ID 计数对结果进行分组，但是，任何尝试按日期对结果进行分组，或将查询包装为派生表并再次查询，都会导致大约 6 分钟后超时。我认为相关子查询可能正在重新运行。我无法在 mysql 5.7 中使用 CTE，并且无法使用初始结果创建临时表。

初始查询 - 运行良好

select
        valid_sellers.uid
        ,(select min(si.created) as 'temp_name' 
            from sell_item si where si.uid = valid_sellers.uid and si.status in ("For Sale", "Sold")  
            having temp_name >= "2023-01-01" and temp_name <= "2023-12-31") as 'first_sales_activity'
from 
    (
        select
            u.uid
        from
            user u
        where
            u.uid in (
                    select
                        t1.uid
                     from
                        (select distinct si.uid from sell_item si) t1)
            and
            u.since < "2023-04-24"
            and
            u.can_sell = "Y"
            and
            u.last_alive_time >= "2023-01-01"
    ) valid_sellers 

尝试聚合会导致超时

select
        count(valid_sellers.uid)
        ,DATE_FORMAT((select min(si.created) as 'temp_name' 
            from sell_item si where si.uid = valid_sellers.uid and si.status in ("For Sale", "Sold")  
            having temp_name >= "2023-01-01" and temp_name <= "2023-12-31"), '%Y-%m') as 'year_month'
from 
    (
        select
            u.uid
        from
            user u
        where
            u.uid in (
                    select
                        t1.uid
                     from
                        (select distinct si.uid from sell_item si) t1)
            and
            u.since < "2023-04-24"
            and
            u.can_sell = "Y"
            and
            u.last_alive_time >= "2023-01-01"
    ) valid_sellers

group by
    2
    

第二次尝试聚合导致超时

select
    count(first_listings.uid)
    ,DATE_FORMAT(first_listings.first_sales_activity, '%Y-%m') as 'first_listing'
    
from (
        select
                valid_sellers.uid
                ,(select min(si.created) as 'temp_name' 
                    from sell_item si where si.uid = valid_sellers.uid and si.status in ("For Sale", "Sold")  
                    having temp_name >= "2023-01-01" and temp_name <= "2023-12-31") as 'first_sales_activity'
        from 
            (
                select
                    u.uid
                from
                    user u
                where
                    u.uid in (
                            select
                                t1.uid
                             from
                                (select distinct si.uid from sell_item si) t1)
                    and
                    u.since < "2023-04-24"
                    and
                    u.can_sell = "Y"
                    and
                    u.last_alive_time >= "2023-01-01"
            ) valid_sellers
    
    ) first_listings 

group by
    2

Answer 1

我不认为 MySQL 很好地实现了相关子查询。改为

JOIN

。

而且，它在优化方面真的很糟糕

WHERE uid IN (subquery)

——这几乎应该总是一个连接。

当您加入子查询时，它会过滤掉该子查询未返回的用户，因此您甚至不需要

WHERE IN

检查。

WITH d AS (
    SELECT uid, MIN(created) AS first_sales_activity
    FROM sell_item
    WHERE status IN ("For Sale", "Sold") AND created BETWEEN '2023-01-01' AND '2023-12-31'
    GROUP BY uid
)

SELECT u.uid, DATE_FORMAT(d.first_sales_activity, '%Y-%m') AS first_sales_activity
FROM users AS u
JOIN d ON u.uid = d.uid
WHERE u.since < '2023-04-014' AND u.can_sell = 'Y' AND u.last_alive_time >= "2023-01-01"