1

我有这个查询：

select '2012-Nov-01' As "Week_Of",
count(player_id) as cohort_size ,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 0 
    then player_id 
  end) as Day_0_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 1 
    then player_id 
  end) as Day_1_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 2
    then player_id 
  end) as Day_2_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 3
    then player_id 
  end) as Day_3_Ret
from player
where trunc(create_dtime) > To_Date('2012-Nov-01','yyyy-mon-dd')-7
and trunc(create_dtime) <= To_Date('2012-Nov-01','yyyy-mon-dd')
and world_flag != '1'
Which outputs:

Week_Of       Cohort_size    Day_0_Ret    Day_1 Ret    Day_2_Ret    Day_3_Ret
2012-Nov-01    2426           2426        1412         1349          1316

我希望能够针对多组日期运行此查询，以便我的输出看起来像：

Week_Of       Cohort_size    Day_0_Ret    Day_1 Ret    Day_2_Ret    Day_3_Ret
2012-Nov-01    2426           2426        1412         1349          1316
2012-Nov-08    5106           2458        1945         1379          1248
2012-Nov-15    3580           1476        1412         1349          1146

而不是在特定的一周内每次都重新运行查询。这可能吗？

4

1 回答 1

4

以下查询将您的逻辑更改为 agroup by而不是 where：

select to_char(max(create_dtime), 'yyyy-mm-dd') As "Week_Of",
       count(player_id) as cohort_size ,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 0 
    then player_id 
  end) as Day_0_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 1 
    then player_id 
  end) as Day_1_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 2
    then player_id 
  end) as Day_2_Ret,
count(case 
    when trunc(init_dtime)-trunc(create_dtime) >= 3
    then player_id 
  end) as Day_3_Ret
from player
where trunc(create_dtime) >= To_Date('2012-Nov-01','yyyy-mon-dd') - 6
and world_flag != '1'
group by trunc((trunc(create_dtime) - (To_Date('2012-Nov-01','yyyy-mon-dd') -  6))/7)
order by 1

它选择最后一个可用的日期来描述一周....

于 2013-03-06T23:26:12.887 回答