在oracle sql中获取每个月的累计总和

嗨，我有一张有三列的桌子

custid  p_date          vegetable
0000009 07-APR-20       tomato
0000013 25-MAR-20       potato
0000015 11-APR-19       tomato
0000016 23-JUL-18       potato
0000019 19-JUL-18       potato
0000018 07-JUN-20       turnip
0000020 12-JUL-18       turnip
0000022 23-JUL-18       potato
0000023 12-JUL-18       turnip
0000024 11-AUG-20       potato

我需要根据蔬菜计算顾客的不同数量 列以及每个月的累计金额

输出应包含以下列

date        
tomato_ct 
cum_tomato_ct
potato_ct
cum_potato_ct
turnip_ct        
cum_turnip_ct

---

如果希望每月计数和累积计数以蔬菜为轴心，可以执行以下操作：

select trunc(p_date, 'month') as p_month,
    sum(case when vegetable = 'tomato' then 1 else 0 end) as tomato_ct,
    sum(sum(case when vegetable = 'tomato' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_tomato_ct,
    sum(case when vegetable = 'potato' then 1 else 0 end) as potato_ct,
    sum(sum(case when vegetable = 'potato' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_potato_ct,
    sum(case when vegetable = 'turnip' then 1 else 0 end) as turnip_ct,
    sum(sum(case when vegetable = 'turnip' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_turnip_ct
from mytable
group by trunc(p_date, 'month')

您提到您需要统计不同客户的数量。如果你想对每一个客户只计算一次蔬菜，在他们最早的外观，那么我会建议两个层次的聚合：

select trunc(p_date, 'month') as p_month,
    sum(case when vegetable = 'tomato' then 1 else 0 end) as tomato_ct,
    sum(sum(case when vegetable = 'tomato' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_tomato_ct,
    sum(case when vegetable = 'potato' then 1 else 0 end) as potato_ct,
    sum(sum(case when vegetable = 'potato' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_potato_ct,
    sum(case when vegetable = 'turnip' then 1 else 0 end) as turnip_ct,
    sum(sum(case when vegetable = 'turnip' then 1 else 0 end)) over(order by trunc(p_date, 'month')) as cum_turnip_ct
from (
    select cust_id, vegetable, min(p_date) as p_date
    from mytable
    group by cust_id, vegetable
) t
group by trunc(p_date, 'month')

---

您可以使用分析函数和分组方式的组合，如下所示：

SQL>
SQL> -- sample data
SQL> WITH YOUR_TABLE(custid,  p_date,          vegetable) AS
  2  (SELECT '0000009', TO_DATE('07-APR-20','DD-MON-RR'),       'tomato' FROM DUAL UNION ALL
  3  SELECT '0000013', TO_DATE('25-MAR-20','DD-MON-RR'),       'potato' FROM DUAL UNION ALL
  4  SELECT '0000015', TO_DATE('11-APR-19','DD-MON-RR'),       'tomato' FROM DUAL UNION ALL
  5  SELECT '0000016', TO_DATE('23-JUL-18','DD-MON-RR'),       'potato' FROM DUAL UNION ALL
  6  SELECT '0000019', TO_DATE('19-JUL-18','DD-MON-RR'),       'potato' FROM DUAL UNION ALL
  7  SELECT '0000018', TO_DATE('07-JUN-20','DD-MON-RR'),       'turnip' FROM DUAL UNION ALL
  8  SELECT '0000020', TO_DATE('12-JUL-18','DD-MON-RR'),       'turnip' FROM DUAL UNION ALL
  9  SELECT '0000022', TO_DATE('23-JUL-18','DD-MON-RR'),       'potato' FROM DUAL UNION ALL
 10  SELECT '0000023', TO_DATE('12-JUL-18','DD-MON-RR'),       'turnip' FROM DUAL UNION ALL
 11  SELECT '0000024', TO_DATE('11-AUG-20','DD-MON-RR'),       'potato' FROM DUAL)
 12  -- actual query starts from here
 13  SELECT P_DATE,
 14     COUNT(CASE WHEN vegetable = 'tomato' THEN 1 END) AS TOMATO_CT,
 15     MAX(TOMATO_CNT) AS CUM_TOMATO_CT,
 16     COUNT(CASE WHEN vegetable = 'potato' THEN 1 END) AS POTATO_CT,
 17     MAX(POTATO_CNT) AS CUM_POTATO_CT,
 18     COUNT(CASE WHEN vegetable = 'turnip' THEN 1 END) AS TURNIP_CT,
 19     MAX(TURNIP_CNT) AS CUM_TURNIP_CT
 20    FROM
 21    (SELECT T.*,
 22   COUNT(CASE WHEN vegetable = 'tomato' THEN 1 END) OVER (ORDER BY P_DATE) AS TOMATO_CNT,
 23   COUNT(CASE WHEN vegetable = 'potato' THEN 1 END) OVER (ORDER BY P_DATE) AS POTATO_CNT,
 24   COUNT(CASE WHEN vegetable = 'turnip' THEN 1 END) OVER (ORDER BY P_DATE) AS TURNIP_CNT
 25  FROM (SELECT DISTINCT custid, TRUNC(p_date,'MON') P_DATE, vegetable FROM YOUR_TABLE) T
 26     )
 27   GROUP BY P_DATE
 28   ORDER BY P_DATE;

P_DATE                TOMATO_CT CUM_TOMATO_CT  POTATO_CT CUM_POTATO_CT  TURNIP_CT CUM_TURNIP_CT
-------------------- ---------- ------------- ---------- ------------- ---------- -------------
01-JUL-2018 00:00:00          0             0          3             3          2             2
01-APR-2019 00:00:00          1             1          0             3          0             2
01-MAR-2020 00:00:00          0             1          1             4          0             2
01-APR-2020 00:00:00          1             2          0             4          0             2
01-JUN-2020 00:00:00          0             2          0             4          1             3
01-AUG-2020 00:00:00          0             2          1             5          0             3

6 rows selected.

SQL>

---

请告诉我们您想要的结果，而不仅仅是列名。我得到了我想要的输出。非常感谢。我不同意！！！这将抛出一个错误，因为聚合函数中不允许使用分析函数。@Popeye:这是另一种方式：聚合函数在分析函数中。是的，你是对的。在同一个查询中，如果我需要总客户计数和累积计数，我将如何编辑？请解释一下上面的查询。maxtomato_cnt如何给出累计和？因为内部查询具有对所有提到的蔬菜的累计计算。通过使用最大值，它使用分组月份所有累计计数的最大值。如果您将单独执行内部查询，那么您将能够看到外部最大查询背后的逻辑