SQL Case 语句子选择内的子选择分组

Posted 2023-03-23

【中文标题】SQL Case 语句子选择内的子选择分组

【英文标题】：SQL Case statement subselect within a grouped by subselect

【发布时间】：2013-03-26 16:44:35

【问题描述】：

我不知道如何修改我的查询以使用 information.infoDate 和 otherDates.auxDate 之间的最新日期。

SELECT state, COUNT(type) as EQUIPMENT, 
       (SELECT COUNT(type) 
        FROM information a 
        WHERE a.state = b.state 
              AND storedprocedures.to_date(infoDate) < (SELECT current date - 5 days 
                                                        FROM sysibm.sysdummy1) 
        GROUP BY state) as IDLE,
        ROUND((SELECT COUNT(type) 
               FROM information a 
               WHERE a.state = b.state 
                     AND storedprocedures.to_date(infoDate) < (SELECT current date - 5 days 
                                                               FROM sysibm.sysdummy1)
               GROUP BY state) / CAST(COUNT(type) as float) * 100, 1) as percent
FROM information b
JOIN validStates 
  ON state = vstate
LEFT JOIN otherDates 
  ON typeid = auxtypeid
WHERE state <> '***'
GROUP BY state
ORDER BY state

我已经尝试过使用以下语句而不是 infoDate 的各种方式，但我无法运行查询。有没有更好的方法来使用两个日期中的最新日期？

SELECT CASE WHEN auxDate > infoDate 
            THEN auxDate 
            ELSE infoDate END AS activityDate 
FROM information 
LEFT JOIN otherDates 
       ON typeid = auxtypeid

【问题讨论】：

我们能否获得一些示例起始数据/所需的结果输出？ DB2 的哪个版本/平台？无论您的目标格式是什么（为什么不存储在日期/时间/时间戳类型中），是否有一个“自反”from_date 函数？

【参考方案1】：

您的 CASE 语句是选择 2 个字段的最新值的最佳方式。为什么查询没有运行？

添加 CASE 语句后...确保您也将 CASE 语句添加到查询的 GROUP BY 部分。

哦，...在您头疼之前 - 您可能需要考虑不使用保留关键字（例如状态、类型等）创建字段。

编辑：哎呀我没有看到 DB2 的标签

【讨论】：

因此，如果您要在 infoDate 所在的位置插入 case 语句，您将如何按 case 语句进行分组？我无法按活动日期分组。

如果您可以使用失败的案例语句回复 SQL，也许我可以提供帮助。起初我没有看到 DB2。在 MS SQL 中，您需要按 CASE 语句进行分组……在 DB2 中，也许您可​​以简单地 GROUP BY 语句的别名……例如 GROUP BY activityDate

【参考方案2】：

SELECT state,
COUNT (type) as EQUIPMENT, 
(SELECT COUNT(type) FROM information a LEFT JOIN otherDates on typeid=auxtypeid WHERE a.state=b.state AND (CASE WHEN otherDate > infoDate THEN storedprocedures.dec_to_date(otherDate) ELSE storedprocedures.dec_to_date(infoDate) END) <= (select current date - 5 days from sysibm.sysdummy1) GROUP BY state) as IDLE,
ROUND((SELECT COUNT(type) FROM information a WHERE a.state=b.state AND (CASE WHEN otherDate > infoDate THEN storedprocedures.dec_to_date(otherDate) ELSE storedprocedures.dec_to_date(infoDate) END) < (select current date - 5 days from sysibm.sysdummy1) GROUP BY state)/ CAST(COUNT(type) as float)*100,1) as percent
FROM information b
JOIN validStates ON state=vstate
LEFT JOIN otherDates ON typeid=auxtypeid
WHERE state<>'***'
GROUP BY state
ORDER BY state;

【参考方案3】：

您可能会更好地执行以下操作（直到我对您的表结构/数据了解更多）：

WITH Raw as (SELECT a.state, 
                    MAX(a.infoDate, COALESCE(c.auxDate, '0001-01-01')) as activityDate
             FROM information a
             JOIN validStates b
               ON b.vstate = a.state
             LEFT JOIN otherDates c
                    ON typeId = auxtypeid
             WHERE a.state <> '***'),
     Total as (SELECT state, COUNT(*) as count
               FROM Raw
               GROUP BY state),
     Idle as (SELECT state, COUNT(*) as count
              FROM Raw
              WHERE storedprocedures.to_date(activityDate) < (CURRENT_DATE - 5 DAYS)
              GROUP BY state)
SELECT a.state, a.count as EQUIPMENT, b.count as IDLE, 
       ROUND((100.0 * b.count) / a.count, 1) as percent
FROM Total a
LEFT JOIN Idle b
       ON b.state = a.state
ORDER BY a.state

...它也可能会运行得更快。

请注意，由于使用了LEFT JOINs，如果给定的state 没有“空闲”类型，IDLE 和percent 可能为空。