需要帮助在Postgres中用窗口函数替换子查询

Question

I have a query that returns the results I want, however it takes 3 full table scans because of the subqueries. My attempts to turn this into window functions have not worked. Thoughts?

SELECT a_day
    ,max(week_count)
    ,max(month_count)
FROM (
    SELECT DISTINCT date_trunc('day', created_at) AS a_day
        ,(
            SELECT count(*)
            FROM accounts AS wk
            WHERE wk.created_at BETWEEN (a.created_at - INTERVAL '7 days')
                    AND a.created_at
            ) AS week_count
        ,(
            SELECT count(*)
            FROM accounts AS wk
            WHERE wk.created_at BETWEEN (a.created_at - INTERVAL '30 days')
                    AND a.created_at
            ) AS month_count
    FROM accounts a
    ORDER BY a_day
    ) AS sub_1
GROUP BY a_day
ORDER BY a_day

Answer 1

with sub_1 as
(select date_trunc('day', created_at) AS a_day, count(*) as week_count
 from accounts
 where created_at BETWEEN (created_at - INTERVAL '7 days') and created_at
 group by date_trunc('day', created_at))
,sub_2 as 
(select date_trunc('day', created_at) AS a_day, count(*) as month_count
 from accounts
 where created_at BETWEEN (created_at - INTERVAL '30 days') and created_at
 group by date_trunc('day', created_at))
 select sub_1.a_day, max(sub_1.week_count), max(sub_2.month_count)
 from sub_1 join sub_2 on sub_1.a_day = sub_2.a_day
 group by sub_1.a_day

Try using the Common table expressions and see if you can get the performance gain. This scans the table 2 times instead of 3.