统计分钟级别的视频在线用户数+列炸裂+repeat函数

统计分钟级别的视频在线用户数

1、原始数据如下:

uid vid starttime endtime

select 'aa' as uid,'v00l' as vid,'2023-10-25 12:00' as starttime,'2023-10-2512:15' as endtime

union

select 'bb' as uid,'v002' as vid,'2023-10-25 12:05' as starttime,'2023-10-25 12:19' as endtime

结果如下:

2、需求分析

  • 将用户看视频的时间拆解到分钟
  • 现在能算出用户看视频的时长(分钟),并且知道用户看视频的开始时间,如果拆解到分钟级别观看记录呢?
  • 肯定涉及到列转行,但是如何拆分成 时长单位(分钟) 个 分钟级别观看记录行?

3、完整代码如下

3.1、实现一:repeat+explode+row_number()over(partitions by )
sql 复制代码
--step6、统计每分钟看视频的用户数,并按照时间排序
select 
    each_online_minute,from_unixtime(each_online_minute,'yyyy-MM-dd hh:mm') as minute_lab bel,count(distinct uid) online_uid_cnt 
from
(--step5、形成每个用户看视频的分钟级别的观看记录
   select
        uid,vid,starttime,starttimestamp, endtime, endtimestamp,min_cnt
        ,row_number()over(part: ition by uid,vid,starttime order by endtimestamp) rk
        ,starttimestamp+60*row_number()over(partition by uid,vid,starttime order by endtimest tamp) as each_online_minute 
   
   from
   (--step4、explode列炸裂
       select uid,vid, starttime,starttimestamp ,endtime,endtimestamp, min_cnt,new_repeat 
       from
       (--step3、用repeat产生持续时长(分钟) 个数组
           select uid,vid,starttime,starttimestamp,endtime,endtimestamp,min_cnt,repeat('a,',mir n_cnt-1) as repeat_str 
           from
           (--step2、转化为时间戳,且计算看视频持续时长(分钟)
              select 
                 uid
                 ,vid
                 ,starttime
                 ,unix_timestamp(starttime,'yyyy-MM-dd hh:mm' ) as starttimestamp
                 ,endtime
                 ,unix_timestamp(endtime, 'yyyy-MM-dd hh:mm' ) as endtimestamp
                 ,(unix_timestamp(endtime,'yyyy-MM-dd hh:mm')-unix_timestamp(starttime,'yyyy-Mr M-dd hh:mm'))/60 as min_cnt 
              from
              (--step1、获取原始数据
                  select 'aa' as uid,'v0ol' as vid,'2023-10-25 12:00' as starttime,'2023-10-2512:15' as endtime 
                  union
                  select 'bb' as uid,'v0o2' as vid,'2023-10-25 12:05' as starttime,'2023-10-25 12::19' as endtime
              )tb_base
           )tb_tmp
       )tb_final
       lateral view explode(split(repeat_str,',')) tb_tmp as new_repeat
   )tb_outer
) tb
group by each_online_minute
order by each_online_minute asc
3.2、实现二: repeat+posexplode
sql 复制代码
--step6、统计每分钟看视频的用户数,并按照时间排序
select 
    each_online_minute,from_unixtime(each_online_minute,'yyyy-MM-dd hh:mm') as minute_lab bel,count(distinct uid) online_uid_cnt 
from
(--step5、形成每个用户看视频的分钟级别的观看记录
   select
        uid,vid,starttime,starttimestamp, endtime, endtimestamp,min_cnt
        ,starttimestamp+60*repeat_pos as each_online_minute 
   
   from
   (--step4、explode列炸裂
       select uid,vid, starttime,starttimestamp ,endtime,endtimestamp, min_cnt,new_repeat,repeat_pos 
       from
       (--step3、用repeat产生持续时长(分钟) 个数组
           select uid,vid,starttime,starttimestamp,endtime,endtimestamp,min_cnt,repeat('a,',mir n_cnt-1) as repeat_str 
           from
           (--step2、转化为时间戳,且计算看视频持续时长(分钟)
              select 
                 uid
                 ,vid
                 ,starttime
                 ,unix_timestamp(starttime,'yyyy-MM-dd hh:mm' ) as starttimestamp
                 ,endtime
                 ,unix_timestamp(endtime, 'yyyy-MM-dd hh:mm' ) as endtimestamp
                 ,(unix_timestamp(endtime,'yyyy-MM-dd hh:mm')-unix_timestamp(starttime,'yyyy-Mr M-dd hh:mm'))/60 as min_cnt 
              from
              (--step1、获取原始数据
                  select 'aa' as uid,'v0ol' as vid,'2023-10-25 12:00' as starttime,'2023-10-2512:15' as endtime 
                  union
                  select 'bb' as uid,'v0o2' as vid,'2023-10-25 12:05' as starttime,'2023-10-25 12::19' as endtime
              )tb_base
           )tb_tmp
       )tb_final
       lateral view posexplode(split(repeat_str,',')) tb_tmp as repeat_pos,new_repeat
   )tb_outer
) tb
group by each_online_minute
order by each_online_minute asc

4、相关hive函数介绍

4.1、reverse返回字符串反序

用法:reverse('foobar')参数类型string,返回值string,例如:

select reverse('foobar')

--返回:raboof

4.2、space返回指定n个空格字符串

用法:space(10)参数类型int,返回值string 例如:

复制代码
select space(10)
--返回:'         '

4.3、repeat返回字符串重复n次后的字符串

用法:repeat('a',10) 例如:

复制代码
select repeat('a',10)
--返回:'aaaaaaaaaa'

4.3、split将字符串按指定分隔符,拆分为数组 用法:

复制代码
select split('a,b,c,d',',')
--返回:["a","b","c","d"]
相关推荐
smilejingwei1 小时前
TRAE 插件上新 | SQLazy:让 AI 写出可信复杂 SQL
数据库·人工智能·sql
我要见SA姐11 小时前
GitHub Actions 自动化:5 个真实运维提效案例
运维·数据库·机器学习·oracle·回归
QXWZ_IA2 小时前
边坡监测预算有限时的建设顺序:虚拟基准站与成本账
网络·数据库
这个DBA有点耶3 小时前
1亿条向量内存降95.4%,延迟降80%:金仓AI原生内核真能解决RAG痛点?
数据库·人工智能·开源
Y3815326623 小时前
用 httpx 写 SerpBase 异步客户端:连接池与超时配置
运维·服务器·数据库
c_hn_0073 小时前
量化数据12:国内期货L2历史数据:Tick、逐笔、五档盘口一次性打包下载
数据库·1024程序员节·#深度订单簿数据·#历史行情订单薄·#本地csv数据存储
GeW3 小时前
从"补短板"到"强优势"Linux与数据库认证成为产业跃迁人才支点
linux·数据库
数据库小学妹4 小时前
MySQL NULL陷阱:NOT IN、三值逻辑与NOT EXISTS
数据库·后端·mysql
SelectDB4 小时前
Doris 异步物化视图实战:与 Snowflake MV / Dynamic Tables 分层建模对照
大数据·数据库·云原生
枫昕柚4 小时前
MySQL 建库命令详解:从 CREATE DATABASE 到create table
数据库·mysql