获取按值分组的百分比值

Posted 2023-03-24

【中文标题】获取按值分组的百分比值

【英文标题】：getting percentage value for grouped by values

【发布时间】：2021-06-21 16:18:08

【问题描述】：

我有一张桌子，我正在使用该组来获取每个国家/地区的分组总和

Select country,count(country) from `xyz.xyz.call_stats` where  fan_out = 1 and last_channel_uri not like 'sip:confctl-d%' and call_end_time between '2021-05-17 00:00:00' and '2021-05-18 00:00:00' group by country

它以以下格式为我提供值：-

Row     country     f0_     
1       DZ          4941
    
2       NO          30737
    
3       IS          436
    
4       IT          32086
    
5       GF          11
    
6       CZ          9156

我需要的是每个国家/地区的百分比值，例如国家 DZ，4941 是该时间段的百分比值，并且所有时间段都相同。我的预期结果是：-

    Row     country     f0_     f1_
    1       DZ          4941    10%
        
    2       NO          30737  25.7%
        
    3       IS          436    2%
        
    4       IT          32086  29.6%
        
    5       GF          11     0.04%
        
    6       CZ          9156   3.67%

P.S:- 我想在计算总百分比时​​忽略条件 fan_out = 1 但在进行分组时想保留它。

【参考方案1】：

您可以使用窗口函数。我更喜欢 0 和 1 之间的比率：

Select country, count(*),
       count(*) * 1.0 / sum(count(*)) over () as ratio
from `xyz.xyz.call_stats`
where fan_out = 1 and
      last_channel_uri not like 'sip:confctl-d%' and
      call_end_time between '2021-05-17 00:00:00' and '2021-05-18 00:00:00'
group by country;

如果你想要一个介于 0 和 100 之间的值，你可以乘以 100。

编辑：

如果您想要基于特定条件的百分比，请使用countif()：

Select country, countif(fan_out = 1),
       count(*) * 1.0 / sum(count(*)) over () as ratio
from `xyz.xyz.call_stats`
where last_channel_uri not like 'sip:confctl-d%' and
      call_end_time between '2021-05-17 00:00:00' and '2021-05-18 00:00:00'
group by country;

【讨论】：

那么，如果我想在计算百分比总数时忽略 fan_out=1 条件，但在进行分组时我想保留它，我该怎么办。我知道这不是问题的一部分，但如果你能帮助我，那就太好了。如果你愿意，我也可以编辑问题。

【参考方案2】：

我能够使用以下解决方案解决查询：-

WITH 
T1 AS (Select country,count(country) as NumA from `xyz.xyz.call_stats` where  fan_out = 1 and last_channel_uri not like 'sip:confctl-d%' and call_end_time between '2021-05-17 00:00:00' and '2021-05-18 00:00:00' group by country), 
T2 AS (Select count(*) as NumB from `xyz.xyz.call_stats` where last_channel_uri not like 'sip:confctl-d%' and call_end_time between '2021-05-17 00:00:00' and '2021-05-18 00:00:00') 
SELECT country,CAST(T1.NumA AS FLOAT64) / CAST(T2.NumB AS FLOAT64) * 100 as PctMetTgt
FROM T1, T2 order by PctMetTgt DESC