如何通过分析获得正确的摘要？答案

【问题标题】：How to get correct summaries with analytics?如何通过分析获得正确的摘要？
【发布时间】：2015-07-29 22:21:28
【问题描述】：

如果特定的invoice_code 出现在invoice_detail 中，我想从cust_detail 表中获取摘要编号。

在此示例中，我只想报告批次 10 和 20 的 cust_detail 摘要，因为它们是带有 invoice_code='9999' 的批次。但是invoice_detail 表中的重复数据扭曲了我的数字。

with
  invoice_detail as
  (
    select '10' as invoice_batch, '9999' as invoice_code from dual union all 
    select '10' as invoice_batch, '9999' as invoice_code from dual union all
    select '20' as invoice_batch, '1111' as invoice_code from dual union all
    select '30' as invoice_batch, '9999' as invoice_code from dual
  ),
  cust_detail as
  (
    select '1' as cust_id, '10' as invoice_batch, 40 as points_paid, 30 as points_earned, 30 as points_delivered from dual union all
    select '1' as cust_id, '20' as invoice_batch, 10 as points_paid, 10 as points_earned, 10 as points_delivered from dual union all
    select '1' as cust_id, '30' as invoice_batch, 20 as points_paid, 15 as points_earned,  5 as points_delivered from dual
  )
select cust_id, 
       sum(points_paid) over (partition by  c.invoice_batch
                                  order by cust_id) batch_total
  from cust_detail c
 inner join invoice_detail i on c.invoice_batch=i.invoice_batch
 where i.invoice_code = '9999';

期望的结果：

 CUST_ID   PAID   EARNED   DELIVERED   TOT_PAID   TOT_EARNED   TOT_DELIVERED
--------- ------ -------- ----------- ---------- ------------ ---------------
 1         40     30       30          60         45           40      
 1         20     15       5           60         45           40

【问题讨论】：

您在查询中缺少group by。
当我在查询末尾添加 group by c.invoice_batch 时，我收到 not a group by 表达式消息。
你也应该group by cust_id
我在这里使用分析函数。按invoice_batch 或cust_id 分组产生ORA-00979: Not a group by expression 消息。我认为group by 不适用于sum() 等分析函数

标签： sql oracle sum window-functions

【解决方案1】：

您可以在加入前使用 distinct 从 invoice_detail 中删除重复项：

with invoice_detail as
(
select '10' as invoice_batch, '9999' as invoice_code from dual union all 
select '10' as invoice_batch, '9999' as invoice_code from dual union all
select '20' as invoice_batch, '1111' as invoice_code from dual union all
select '30' as invoice_batch, '9999' as invoice_code from dual
),
cust_detail as
(
select '1' as cust_id, '10' as invoice_batch, 40 as points_paid, 30 as points_earned, 30 as points_delivered from dual union all
select '1' as cust_id, '20' as invoice_batch, 10 as points_paid, 10 as points_earned, 10 as points_delivered from dual union all
select '1' as cust_id, '30' as invoice_batch, 20 as points_paid, 15 as points_earned,  5  as points_delivered from dual
)
select cust_id
            ,points_paid
            ,points_earned
            ,points_delivered
            ,sum(points_paid) over (partition by c.cust_id) as tot_paid
            ,sum(points_earned) over (partition by c.cust_id) as tot_earned
            ,sum(points_delivered) over (partition by c.cust_id) as tot_delivered           
from cust_detail c 
join (select distinct * from invoice_detail) i
     on c.invoice_batch=i.invoice_batch
where i.invoice_code = '9999';

请注意，摘要包括批次 10 和 30，因为批次 20 的 invoice_code='1111'。

SQL Fiddle

【讨论】：

【解决方案2】：

我不确定您想要的结果与您的查询有什么关系。但是，我希望您的查询看起来更像这样：

select cust_id, 
       sum(points_paid) over (partition by cust_id) as batch_total
from cust_detail c inner join
     invoice_detail i
     on c.invoice_batch=i.invoice_batch
where i.invoice_code = '9999' ;

【讨论】：