次日留存率

--用户数据表结构
--yyyymmdd 代表用户测试这句话说的日期 ---dt
--nlu_utterance 代表用户说的某句话   ---platform
--svicd 代表用户的手机码 ------ mid


--sf_user_data--代表用户说某句话的所有log
--sf_first_user_data 所有的新增用户




--表1、记录用户首次使用的日期


drop table if exists cincinnati_com.sf_first_user_data
create table if not exists cincinnati_com.sf_first_user_data as


--SELECT *, Row_Number() OVER (partition by deptid ORDER BY salary desc) rank FROM employee


select * ,Row_Number () OVER (partition by table1.svicd ORDER BY table1.uid esc) rank from(
select svicd,
       uid,
  nlu_utterance
from kpi_team.bixby_daily_report
where language = 'zh-CN'
and yyyymmdd between 20180208 and 20180228
and nlu_utterance like '%XX%'
) as table1 
where rank = 1

------------------------------------------------------
--表2、创建该表让新增用户时间+1 和 sf_user_data 做个inner join ，取 其 的数量 / sf_first_user_data 的数量 为 次日留存率
select 
   svicd,
   uid,
   nlu_utterance
 from
(select 
svicd,
uid,
    nlu_utterance,
    yyyymmdd
from sf_first_user_data
as table1)
where yyyymmdd = table1.yyyymmdd






--日期增加函数
 --date_add(string startdate, int days)
 
 select
    date_add('yyyymmdd',1) from test;

 
 
































insert into report_userinfo   
select   
xinzeng.platform,  
xinzeng.dt,  
xinzeng.channel,  
xinzeng.cver,  
2,  
count(distinct cunliu.pcid),  
datediff(  
    from_unixtime(unix_timestamp(cast(cunliu.dt as string),'yyyyMMdd')),  
    from_unixtime(unix_timestamp(cast(xinzeng.dt as string),'yyyyMMdd'))  
)  
from  
(  
    select * from user_login_history where type=1  
) xinzeng  
inner join   
(  
    select   
    platform,  
    dt,  
    channel,  
    cver,  
    mid pcid   
    from log_vvim   
    where mid is not null and dt=$dt group by platform,dt,channel,cver,mid  
) cunliu on  
(  
    xinzeng.platform=cunliu.platform and  
    xinzeng.channel=cunliu.channel and  
    xinzeng.cver=cunliu.cver and  
    xinzeng.pcid=cunliu.pcid  
)  
where cunliu.dt>xinzeng.dt   
group by   
xinzeng.platform,xinzeng.dt,xinzeng.channel,xinzeng.cver,  
datediff(  
    from_unixtime(unix_timestamp(cast(cunliu.dt as string),'yyyyMMdd')),  
    from_unixtime(unix_timestamp(cast(xinzeng.dt as string),'yyyyMMdd'))  
);  










select  retention_sum(xwho_state, 4, 30, 1, 5) 
from (
select  retention_count(
date_diff('day', from_iso8601_timestamp('2018-02-26'), to_timestamp(day, 'yyyymmdd')), 4, 1, 5, 
case when ( day >= '20180226' and day <= '20180301'  and standard_event_id = 1311 and standard_app_id=100361  ) and ( day >= '20180227' and day <= '20180331'  and standard_event_id = 132678 and standard_app_id=100361  ) then 0  
when ( day >= '20180226' and day <= '20180301'  and standard_event_id = 1311 and standard_app_id=100361  ) then 1 else -1 end  , 1, -1) as xwho_state
from fz_bds.bds_ait_device_event_info_d eventAttr 
where
        (
         (day >= '20180226' and day <= '20180301'  and standard_event_id = 1311 and standard_app_id=100361 ) or
     (day >= '20180227' and day <= '20180331'  and standard_event_id = 132678 and standard_app_id=100361 )
    )
    
group by  eventAttr.tmp_id
)