我想使用sql的开始时间和结束时间在presto上统计在线人数。
我的数据如下:
userid begintime endtime
023150000030040 2020-03-05 12:50:46 2020-03-05 12:50:49
023150004186637 2020-03-05 10:31:19 2020-03-05 10:31:24
023150000788581 2020-03-05 00:59:01 2020-03-05 01:02:00
023150004411606 2020-03-05 19:55:42 2020-03-05 20:02:51
023150004066308 2020-03-05 18:48:03 2020-03-05 18:58:03
023150002033547 2020-03-05 12:39:24 2020-03-05 12:42:21
023150000030040 2020-03-05 13:26:02 2020-03-05 13:26:04
023150003690798 2020-03-05 02:04:50 2020-03-05 02:14:50
023150000030040 2020-03-05 13:57:10 2020-03-05 13:57:12
023150004460558 2020-03-05 16:44:48 2020-03-05 16:47:58我想每小时统计一次在线人数。现在我有了一种愚蠢的数数方法。我的sql如下:
select '01' as hour,COUNT(distinct T.userid)
from datamart_ott_b2b_jsydcp.f_tplay t where t.topicdate ='2020-03-05'
and t.begintime < date_parse('2020-03-05 01', '%Y-%m-%d %h')
and t.endtime > date_parse('2020-03-05 00', '%Y-%m-%d %h')
union all
select '02' as hour,COUNT(distinct T.userid)
from datamart_ott_b2b_jsydcp.f_tplay t where t.topicdate ='2020-03-05'
and t.begintime < date_parse('2020-03-05 02', '%Y-%m-%d %h')
and t.endtime > date_parse('2020-03-05 01', '%Y-%m-%d %h')
.......有没有更简单的方法来做这件事?THX
发布于 2020-03-08 20:09:07
在Prestodb中,您可以生成一个包含整数值的数组,然后对它们进行解嵌套以获得小时数。然后使用joins和group by执行所需的计算:
select hh.hh as hour, cont(distinct t.userid)
from (select sequence(0, 23) hhs
) h cross join
unnest(h.hhs) as hh(hh) left join
datamart_ott_b2b_jsydcp.f_tplay t
on hour(begintime) <= hh.hh and
hour(enddtime) >= hh.hh
where t.topicdate = '2020-03-05'
group by hh.hh
order by hh.hh;发布于 2020-03-08 18:01:16
使用日历表方法,我们可以在小时和日期都匹配的条件下,将包含所有24小时的表左连接到当前表。然后,我们可以按小时汇总,并计算不同用户的数量,以生成您想要的输出。
WITH hours AS (
SELECT 0 AS hour UNION ALL
SELECT 1 UNION ALL
SELECT 2 UNION ALL
...
SELECT 23
)
SELECT
h.hour,
COUNT(DISTINCT t.userid) AS user_cnt
FROM hours h
LEFT JOIN datamart_ott_b2b_jsydcp.f_tplay t
ON h.hour = DATE_TRUNC('hour', t.topicdate) AND
t.topicdate = '2020-03-05'
GROUP BY
h.hour
ORDER BY
h.hour;https://stackoverflow.com/questions/60586271
复制相似问题