按条件小于分组列的clickhouse分组

Question

我有部分sql代码：

select   
    ,inst_date
    ,country
    ,count (distinct (case when (event_day - inst_date) <= 0 then id end)) as event_0
    ,count (distinct (case when (event_day - inst_date) <= 1 then id end)) as event_1
    ,count (distinct (case when (event_day - inst_date) <= 2 then id end)) as event_2
    ,count (distinct (case when (event_day - inst_date) <= 3 then id end)) as event_3
    ,count (distinct (case when (event_day - inst_date) <= 4 then id end)) as event_4
    ,count (distinct (case when (event_day - inst_date) <= 5 then id end)) as event_5
    ,count (distinct (case when (event_day - inst_date) <= 6 then id end)) as event_6
    ,count (distinct (case when (event_day - inst_date) <= 7 then id end)) as event_7
from t1
all left join 
(
    select distinct id
         , toDate(event_date) as event_day
         , 1 as event
    from events
) as events_
using id
where (event_day - inst_date) between 0 and 7
group by inst_date
    ,country ;

我看到了我的数据：

inst_date	country	event_0	event_1	event_2	event_3	event_4	event_5	event_6	event_7
2020-12-01	us	10	11	11	12	13	14	14	14

但我想像这样收集这些数据

inst_date	country	date_difference	events_count
2020-12-01	us	0	10
2020-12-01	us	1	11
2020-12-01	us	2	11
2020-12-01	us	3	12
2020-12-01	us	4	13
2020-12-01	us	5	14
2020-12-01	us	6	14
2020-12-01	us	7	14

在 t1 表 - 带有国家和 inst 日期的 id 列表。在事件表 - 按 id 的事件日期

Answer 1

使用group by:

select inst_date, country, 
    (event_day - inst_date) as date_difference,
    count(*) as events_count
from t1
all left join (
    select distinct id, toDate(event_date) as event_day, 1 as event
    from events
) as events_ using id
where (event_day - inst_date) between 0 and 7
group by inst_date, country, (event_day - inst_date)
order by inst_date, country, date_difference;

我强烈建议在查询中的所有列之前加上它们所属的表，这样查询就可以明确地了解底层数据结构。