，Postgres的查询 - 找到开始和周结束之差，每周

Question

我有被发送daily领域

total_user_count, sent_at

列user_report。

我试图找到total new users per week，其被定义为

difference in user count from the beginning of the week to the ending of the week.

我似乎遇到了问题，虽然制定此查询，我有什么是

SELECT DATE_TRUNC('day', sent_at AT TIME ZONE '+08') as date, total_user_count 
FROM user_state_report 
WHERE timestamptz_cmp(DATE_TRUNC('week', sent_at AT TIME ZONE '+08') , DATE_TRUNC('day', sent_at AT TIME ZONE '+08'))=0 
ORDER BY TIMESTAMP 

哪给我每周的总用户数。但我想每周新注册，而不是总数量

理想输出：

Date [monday of the week X], users registered in this week 14-08-2017 300 21-08-2017 400

谢谢！ :)

Answer 1

你应该能够用滞后函数做到这一点。喜欢的东西：

SELECT date, total_user_count - COALESCE(LAG(total_user_count) OVER(order by date),0) AS new_users_registered from 
(SELECT DATE_TRUNC('day', sent_at AT TIME ZONE '+08') as date, total_user_count 
FROM user_state_report 
WHERE timestamptz_cmp(DATE_TRUNC('week', sent_at AT TIME ZONE '+08') , DATE_TRUNC('day', sent_at AT TIME ZONE '+08'))=0 
ORDER BY sent_at) l 

BTW我改变了你的ORDER BY TIMESTAMP到ORDER BY sent_at，因为我没有undertstand如何你一个工作！