REDSHIFT：SIZE 低于 65K 但“错误：结果大小超过 LISTAGG 限制代码：8001”

Question

Redhisft listagg DISTINCT function 上是否有可能存在错误？

当我尝试这个

select 
case when   partition_field IS NOT NULL  THEN regexp_count(
          listaggdistinct(trim(counted_field), ',')  over (partition by partition_field ), 
        ','
        ) + 1 END AS counted_distinct
from 
TABLE

我收到消息错误：结果大小超出 LISTAGG 限制代码：8001

但是我正在连接一个最大长度为 18 的 varchar 字段，此查询验证了该字段

select max(length(counted_field)) from TABLE

如果我运行以下命令，我会得到 10...

select max(COUNTER) from 
    (SELECT  case when   partition_field IS NOT NULL THEN
         count(  counted_field) over (partition by partition_field )
       END as COUNTER
        from TABLE
       )x

所以 10*18... 比 65K 小很多！！！

我究竟做错了什么？

请注意，我与此代码 100% 与 Redshift 文档一致的结果完全相同

select 
case when   partition_field IS NOT NULL  THEN regexp_count(
          listagg( distinct counted_field, ',')  within group (order by counted_field) over (partition by partition_field ), 
        ','
        ) + 1 END AS counted_distinct
from 
TABLE

在此代码中within group (order by contact_id)只是因为否则Redshift会返回WINDOW错误

Answer 1

答案是 partition_field 可以是 null 并且为了避免计算 partition_field NULL 代码必须如下：

select 
case  regexp_count(
          listaggdistinct ( 
              CASE WHEN   partition_field IS NOT NULL  
              THEN counted_field END , ','
                         )  over (partition by partition_field ), 
        ','
        ) + 1 END AS counted_distinct
from 
TABLE