移动分发端 基础统计指标经典业务代码节选--二次激活用户
二次激活的俩种定义：（以下的实现主要是hiveQL实现）
第一种定义
--二次激活 在首次登陆日期后的30天内有登陆的用户
from log_db.login_etl
insert overwrite table temp_u_statusers_sec partition(td=1)
select firstlogindate statdate,${period} period,softid,platform,firstchannelid
,count(distinct imei) NewUserCount_SecAct,1 flag
where dt>='${datestart}' and dt<='${dateend}' and logindate!=firstlogindate
and firstlogindate>=${firstlogindate} and firstlogindate<=${lastlogindate}
and flag&1=0
group by firstlogindate,softid,platform,firstchannelid
insert overwrite table temp_u_statusers_sec partition(td=2)
select firstlogindate statdate,${period} period,softid,platform,0 channelid
,count(distinct imei) NewUserCount_SecAct,1 flag
where dt>='${datestart}' and dt<='${dateend}' and logindate!=firstlogindate
and firstlogindate>=${firstlogindate} and firstlogindate<=${lastlogindate}
and flag&1=0
group by firstlogindate,softid,platform;
--第二种定义
--过去30天内用户在统计日激活用户数
--百度二次激活,1获取新增用户，2筛选出该端时间内没有登陆的用户 3，没有登陆用户在统计时间登陆用户
insert overwrite table temp_u_statusers_sec2
select c.softid,c.platform,c.channelid,count(1) newusercount_sec2 from (
select softid,platform,firstchannelid channelid,imei,count(distinct logindate) s
from log_db.login_etl
where dt>='${datestart}' and dt<='${dateend2}' and firstlogindate>=${firstlogindate}
and firstlogindate<=${lastlogindate}
group by softid,platform,firstchannelid,imei
having count(distinct logindate)=1
) c inner join
(
select softid,platform,firstchannelid channelid,imei
from log_db.login_etl
where dt='${dir}' and logindate!=firstlogindate and firstlogindate>=${firstlogindate}
and firstlogindate<=${lastlogindate}
group by softid,platform,firstchannelid,imei
) d
on c.softid=d.softid and c.platform=d.platform and c.imei=d.imei
group by c.softid,c.platform,c.channelid;
个人感觉第二种的写法有偷巧的意思，想比于所写的思路来看，效率更好也更高