Question

运行以下查询时，我遇到了Amazon Redshift聚合错误的问题：

select case when frequency between (avg(frequency) + stddev(frequency)) and (avg(frequency) - stddev(frequency)) then  round(avg(frequency) - stddev(frequency))||'-'||round(avg(frequency) + stddev(frequency))
       when frequency between (avg(frequency) + 2*stddev(frequency)) and (avg(frequency) - 2*stddev(frequency)) then  round(avg(frequency) - 2*stddev(frequency))||'-'||round(avg(frequency) + 2*stddev(frequency))
       when frequency between (avg(frequency) + 3*stddev(frequency)) and (avg(frequency) - 3*stddev(frequency)) then  round(avg(frequency) - 3*stddev(frequency))||'-'||round(avg(frequency) + 3*stddev(frequency))
          else null
           end as deviation 
from schema.table

错误告诉我，我需要在group by子句中包含频率。如果我这样做，那么我会收到“群组中不允许的聚合”。有谁知道为什么会这样吗？我最初的猜测是它可能是数据类型的问题，但是弄乱这个并没有帮助。

谢谢！

Answer 1

这些查询可能会令人困惑，您可以在子查询中单独获取聚合，然后通过交叉连接在每一行上使用它们，或者您可以使用分析函数，这样您就可以获得聚合值而无需GROUP BY：

SELECT case when frequency between (avg_Freq + dev_Freq) and (avg_Freq - dev_Freq) then  round(avg_Freq - dev_Freq)||'-'||round(avg_Freq + dev_Freq)
       when frequency between (avg_Freq + 2*dev_Freq) and (avg_Freq - 2*dev_Freq) then  round(avg_Freq - 2*dev_Freq)||'-'||round(avg_Freq + 2*dev_Freq)
       when frequency between (avg_Freq + 3*dev_Freq) and (avg_Freq - 3*dev_Freq) then  round(avg_Freq - 3*dev_Freq)||'-'||round(avg_Freq + 3*dev_Freq)
          else null
           end as deviation 
FROM schema.table
CROSS JOIN (SELECT avg(frequency) AS avg_Freq
            ,stddev(frequency) AS dev_Freq
      FROM schema.table
      )sub

或者，您可以将OVER()添加到现有查询中的每个聚合中：

select case when frequency between (avg(frequency) OVER() + stddev(frequency) OVER()) and (avg(frequency) OVER() - stddev(frequency) OVER()) then  round(avg(frequency) OVER() - stddev(frequency) OVER())||'-'||round(avg(frequency) OVER() + stddev(frequency) OVER())
       when frequency between (avg(frequency) OVER() + 2*stddev(frequency) OVER()) and (avg(frequency) OVER() - 2*stddev(frequency) OVER()) then  round(avg(frequency) OVER() - 2*stddev(frequency) OVER())||'-'||round(avg(frequency) OVER() + 2*stddev(frequency) OVER())
       when frequency between (avg(frequency) OVER() + 3*stddev(frequency) OVER()) and (avg(frequency) OVER() - 3*stddev(frequency) OVER()) then  round(avg(frequency) OVER() - 3*stddev(frequency) OVER())||'-'||round(avg(frequency) OVER() + 3*stddev(frequency) OVER())
          else null
           end as deviation 
from schema.table

对于redshift语法不是100％，但相信两者都应该有用。

Answer 2

您可以通过以下方式将其分解为：

WITH
SELECT avg(frequency) as AVG, stddev(frequency) as STDDEV 
  from schema.table AS TEMP
,
SELECT case when frequency between TEMP.AVG and TEMP.STDDEV etc.

您必须检查确切的陈述。我是靠头做的。

非聚合不存在时聚合列的问题

2 个答案: