如何使用分析函数逐行计算配额？

Question

我们在 Oracle 中使用分析函数时遇到问题。

我们想从其他列中计算剩余配额列。我们的数据按 branch_id 和 region_id 分组。每个分支和区域都有一个配额值（例如 1077）。 如果配额值大于剩余配额值，我们不会更改剩余配额值。

我希望以下示例具有描述性。

我们创建了一个查询来计算剩余配额的总和，但我们失败了。

示例查询和结果

SELECT   CUSTOMER_NUM, BRANCH_ID, REGION_ID, TOTAL_QUOTA, QUOTA,
                 TOTAL_QUOTA - 
                 SUM (CASE
                           WHEN TOTAL_QUOTA - QUOTA < 0
                              THEN 0
                           ELSE QUOTA
                        END) OVER (PARTITION BY BRANCH_ID, REGION_ID ORDER BY BRANCH_ID,
                    REGION_ID,DESC)      
                     AS REMAINING_QUOTA                     
            FROM MY_TABLE
           WHERE BRANCH_ID = 723 AND REGION_ID = 822
        ORDER BY BRANCH_ID,
                 REGION_ID DESC

请帮助我们解决这个问题。

谢谢...

改进：

事实上，由于查询复杂性降低，我从表中丢弃了一些列。

我们的测试表有这些列。

  CUSTOMER_NUM        NUMBER,
  BRANCH_ID           NUMBER,
  ACCOUNT_NUM         NUMBER,
  TOTAL_QUOTA         NUMBER,
  QUOTA               NUMBER,
  REGION_ID           NUMBER,
  LOB_CODE            NUMBER,
  ESTIMATED_END_DATE  DATE,
  TOTAL_AMOUNT        NUMBER

我们在 SQLFiddle 中创建该表的数据结构。链接：http://sqlfiddle.com/#!4/94fe4/1/0

我原来的失败查询是：

SELECT   CUSTOMER_NUM, BRANCH_ID, REGION_ID,LOB_CODE,ESTIMATED_END_DATE,TOTAL_AMOUNT, TOTAL_QUOTA, QUOTA,
                 TOTAL_QUOTA -
                 SUM (CASE
                           WHEN TOTAL_QUOTA - QUOTA < 0
                              THEN 0
                           ELSE QUOTA
                        END) OVER (PARTITION BY BRANCH_ID, REGION_ID, LOB_CODE ORDER BY BRANCH_ID,
                    REGION_ID,
                    LOB_CODE,
                    ESTIMATED_END_DATE,
                    TOTAL_AMOUNT DESC)  
                     AS REMAINING_QUOTA              
            FROM TEST_TABLE
           WHERE BRANCH_ID = 723 AND REGION_ID = 822
                 AND LOB_CODE = 934
        ORDER BY BRANCH_ID,
                 REGION_ID,
                 LOB_CODE,
                 ESTIMATED_END_DATE,
                 TOTAL_AMOUNT DESC

我修改了您的解决方案查询，其中包括 GREATEST 功能。但这对我们来说还不够。如果剩余配额为负数，则使用此函数将此值设置为零。但是，如果剩余配额为负数，我们不想减少配额值。该值必须保持不变。

我们想得到如下查询的结果：（我们为正确的结果添加“or QUOTA between 471 and 472”行。实际上没有这样的行）

SELECT   CUSTOMER_NUM, BRANCH_ID, REGION_ID,LOB_CODE,ESTIMATED_END_DATE,TOTAL_AMOUNT, TOTAL_QUOTA, QUOTA,
                 TOTAL_QUOTA -
                 SUM (CASE
                           WHEN TOTAL_QUOTA - QUOTA < 0
                      or QUOTA between 471 and 472
                              THEN 0
                           ELSE QUOTA
                        END) OVER (PARTITION BY BRANCH_ID, REGION_ID, LOB_CODE ORDER BY BRANCH_ID,
                    REGION_ID,
                    LOB_CODE,
                    ESTIMATED_END_DATE,
                    TOTAL_AMOUNT DESC)  
                     AS REMAINING_QUOTA              
            FROM TEST_TABLE
           WHERE BRANCH_ID = 723 AND REGION_ID = 822
                 AND LOB_CODE = 934
        ORDER BY BRANCH_ID,
                 REGION_ID,
                 LOB_CODE,
                 ESTIMATED_END_DATE,
                 TOTAL_AMOUNT DESC

Answer 1

看起来基本查询是：

SELECT CUSTOMER_NUM, BRANCH_ID, REGION_ID, TOTAL_QUOTA, QUOTA,
       TOTAL_QUOTA - SUM(QUOTA) OVER (PARTITION BY BRANCH_ID, REGION_ID
                                      ORDER BY BRANCH_ID, REGION_ID DESC
                                     ) AS REMAINING_QUOTA
FROM MY_TABLE
WHERE BRANCH_ID = 723 AND REGION_ID = 822
ORDER BY BRANCH_ID, REGION_ID DESC;

请注意，order by 子句不稳定，因为存在重复。最好有一些其他列，例如 id 或 datetime 值，它们可以真正使排序保持一致。

您似乎担心剩余配额为负数。在计算之后处理。只需使用greatest()函数：

SELECT CUSTOMER_NUM, BRANCH_ID, REGION_ID, TOTAL_QUOTA, QUOTA,
       GREATEST(0,
                TOTAL_QUOTA - SUM(QUOTA) OVER (PARTITION BY BRANCH_ID, REGION_ID
                                               ORDER BY BRANCH_ID, REGION_ID DESC
                                              )
               ) AS REMAINING_QUOTA
FROM MY_TABLE
WHERE BRANCH_ID = 723 AND REGION_ID = 822
ORDER BY BRANCH_ID, REGION_ID DESC;