Question

我有一个表格，将人们输入体重的日期存储到我们的应用程序中，我需要弄清楚如何（或者是否）我可以编写一个查询来确定他们称重的连续周数。

有许多资源提供了一般如何计算条纹的示例，但每周部分让我感到非常沮丧。老实说，我甚至不确定从哪里开始。我的伪代码看起来像是：“计算连续几周的数量至少有1个称重”，但是连续几周的部分很棘手。

有没有人实现过这样的东西？

PostgreSQL 9.3是我正在使用的。称量日期为TIMESTAMP WITHOUT TIME ZONE类型。

Answer 1

实际上你问过一个立方体和sql-query这不是一个好主意但是

select log_id, 
    sum(
    case when "dt" between '2012-01-01' and '2012-01-01'+interval'1 month' and "dr">=user.registration_date::date and exists(select 1 from user.activity ph where ph.user_id = log_id and ph.dt::date  between '2012-01-01' and '2012-01-01'+interval'1 month' ) then 1 
    when "dt" between '2012-01-01' and '2012-01-01'+interval'1 month'  and "dt">=user.registration_date::date and not exists(select 1 from user.activity ph where ph.user_id = log_id and ph.dt::date between '2012-01-01' and '2012-01-01'+interval'1 month' ) then 0
    when "dt" between '2012-01-01' and '2012-01-01'+interval'1 month' and "dt"< user.registration_date::date then -1 else null end) as "2012-01",
    sum(case when "dt" between '2012-02-01' and '2012-02-01'+interval'1 month' and "dt">=user.registration_date::date and exists(select 1 from user.activity ph where ph.user_id = log_id and ph.dt::date between '2012-02-01' and '2012-02-01'+interval'1 month' ) then 1
    when "dt" between '2012-02-01' and '2012-02-01'+interval'1 month'  and "dt">=user.registration_date::date and 
                     not exists(select 1 from user.activity ph where ph.user_id = log_id and ph.dt::date between '2012-02-01' and '2012-02-01'+interval'1 month' ) then 0
    when "dt" between '2012-02-01' and '2012-02-01'+interval'1 month'  and "dt"<user.registration_date::date then -1 else null end) as "2012-02"
            from user.activity where dt between '2012-01-01' and '2012-03-01'  group by log_id

按月计算的用户活动示例矩阵 Value >0 - 用户已在所选月份注册并拥有活动 Value =0用户已注册，但在所选月份内没有活动 Value < 0用户尚未在所选月份注册

计算SQL中的每周条纹

1 个答案: