Question

我有一个表Events，其中有2列dt（日期）和event。我正在尝试获取每天的新事件计数。例如如果事件在今天和明天出现，则应仅在今天（因为这是第一次显示）进行计数，并且不应将其从明天的计数中排除。

+------------+-------+
| dt         | event |
+------------+-------+
| 2019-02-01 | u1    |
| 2019-02-01 | u2    |
| 2019-02-01 | u3    |
| 2019-02-01 | u4    |
| 2019-02-01 | u1    |
| 2019-02-02 | u5    |
| 2019-02-02 | u1    |
| 2019-02-02 | u6    |
| 2019-02-02 | u4    |
| 2019-02-03 | u8    |
| 2019-02-03 | u9    |
| 2019-02-03 | u6    |
+------------+-------+

我正在尝试获得以下输出：

+------------+-------------------------+
| dt         | count(distinct c.event) |
+------------+-------------------------+
| 2019-02-01 |                       4 |
| 2019-02-02 |                       2 |
| 2019-02-03 |                       2 |
+------------+-------------------------+

我正在使用的查询是：

with cte (dt,event,rnk) as
(
select dt, 
event,
row_number() over (partition by dt order by event) as rnk
from events order by event
)
select e.dt, count(distinct c.event) from events e left join cte c
on  e.event <> c.event where e.dt = c.dt
group by 1;

但是我无法过滤掉前一天已经出现的事件。我相信缺少一些小东西。我的查询结果是：

+------------+-------------------------+
| dt         | count(distinct c.event) |
+------------+-------------------------+
| 2019-02-01 |                       4 |
| 2019-02-02 |                       4 |
| 2019-02-03 |                       3 |
+------------+-------------------------+

我不确定LEAD / LAG功能是否可以帮助解决问题。还有一种无需使用任何联接就可以实现此目的的方法。

表创建和插入查询：

create table events (dt Date, event varchar(5));

insert into events values('2019-02-01', 'u1');
insert into events values('2019-02-01',  'u2');
insert into events values('2019-02-01',  'u3');
insert into events values('2019-02-01',  'u4');
insert into events values('2019-02-01',  'u1');
insert into events values('2019-02-02',  'u5');
insert into events values('2019-02-02',  'u1');
insert into events values('2019-02-02',  'u6');
insert into events values('2019-02-02',  'u4');
insert into events values('2019-02-03',  'u8');
insert into events values('2019-02-03',  'u9');
insert into events values('2019-02-03',  'u6');

Answer 1

可以使用row_number找到时是第一天的事件出现。

SQL DEMO

WITH ranks as (
    SELECT *, row_number() over (partition by event order by dt) as rn
    FROM events
) 
SELECT dt, COUNT(event)
FROM ranks
WHERE rn = 1 
GROUP BY dt

输出

|         dt | count |
|------------|-------|
| 2019-02-01 |     4 |
| 2019-02-02 |     2 |
| 2019-02-03 |     2 |

Answer 2

这是旧版本的解决方案：https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=12ba0196cf90994d3c9d853ac62405fe

await asyncCall()

CREATE TABLE events
    (`dt` varchar(10), `event` varchar(2))

INSERT INTO events
    (`dt`, `event`)
VALUES
    ('2019-02-01', 'u1'),
    ('2019-02-01', 'u2'),
    ('2019-02-01', 'u3'),
    ('2019-02-01', 'u4'),
    ('2019-02-01', 'u1'),
    ('2019-02-02', 'u5'),
    ('2019-02-02', 'u1'),
    ('2019-02-02', 'u6'),
    ('2019-02-02', 'u4'),
    ('2019-02-03', 'u8'),
    ('2019-02-03', 'u9'),
    ('2019-02-03', 'u6')

实际上，子查询就是您所需要的；其余的可以在应用程序代码中处理。

每天都有新事件计数

2 个答案: