SQL 分组的结果与预期不符

Posted 2023-05-09

【中文标题】SQL 分组的结果与预期不符

【英文标题】：Results of SQL grouping not as intended

【发布时间】：2015-05-27 13:04:12

【问题描述】：

介绍详情：

我在使用客户端 12c 查询 Oracle SQL Developer 时遇到问题。我研究了关于 SO 的其他几个问题，并在 Google 上搜索了这个问题的答案，但最终所有的答案都是在 GROUP BY 子句中包含所有没有聚合函数的列。

我想要的是每年、每周和每天（分别为FD.YEAR、FD.WEEK 和 FD.DT）为每个类别 (PG.CAT) 获得一个结果。我想总结单位、小时数、错误 (GE.QTY) 和总小时数。我还对两列执行乘法和除法，并连接最多四个其他表。

查询：

`SELECT 
  FD.YEAR, 
  FD.WEEK, 
  PG.DT, 
  PG.CAT, 
  SUM(PG.UNITS) AS UNITS, 
  SUM(PG.HOURS) AS HOURS, 
  PM.MTM,
  PM.MTM * PG.HOURS AS ADJ_MTM,
  (PG.UNITS / (PM.MTM * PG.HOURS)) AS PERC_STANDARD,
  SUM(CASE WHEN GE.QTY IS NULL THEN 0 ELSE GE.QTY END) AS QTY,
  SUM(WH.TOTALHOURS) AS TOTALHOURS 
FROM 
  PROD_GUNS PG 
INNER JOIN PROD_MTM PM 
  ON PG.CAT = PM.CATEGORY 
  AND PM.DEPTNO = '018' 
  AND PG.DT BETWEEN PM.START_DT AND PM.END_DT 
INNER JOIN FISCAL_DATES_DAYS FD 
  ON PG.DT = FD.DT 
LEFT OUTER JOIN PROD_GUNS_ERRORS GE 
  ON PG.EID = GE.EID 
  AND PG.DT = GE.DT 
INNER JOIN WEEKLY_HOURS WH
  ON FD.WEEK = WH.DT_WEEK
  AND FD.YEAR = WH.DT_YEAR
  AND PG.EID = WH.EEXX31
GROUP BY 
  FD.YEAR, 
  FD.WEEK, 
  PG.DT, 
  PG.CAT, 
  PM.MTM,
  PM.MTM * PG.HOURS,
  (PG.UNITS / ( PM.MTM * PG.HOURS)) 
HAVING 
  FD.YEAR = '2015' 
  AND FD.WEEK = '1' 
  AND PG.DT = '29-DEC-14' 
  AND PG.CAT = 'Picking'
ORDER BY 
  PG.DT;`

实际结果：

2015    1   29-DEC-14   Picking 46  0.5  68 34      1.35294117647058823529411764705882352941    0   32.21
2015    1   29-DEC-14   Picking 831 7.72 68 524.96  1.58297775068576653459311185614142029869    0   29.35

预期结果：

2015    1   20-Dec-14   Picking 877 8.22 68 558.96   1.21654501216545     0     61.59

问题：

使用我上面的聚合和分组，为什么这不能给我预期的结果？提前感谢大家提供的任何指导。

【问题讨论】：

如果您只是按列分组，而不是表达式（计算）？

这样做会导致以下错误ORA-00979: not a GROUP BY expression 00979. 00000 - "not a GROUP BY expression"

这就是它如此令人沮丧的原因。所有其他问题/答案状态为在 GROUP BY 中包含没有聚合的所有列。好吧，这样做仍然会返回一个抱怨 GROUP BY 子句的错误。

group by 子句的全部意义在于它告诉查询为 grouped by 字段中的每个唯一值组合返回一个单独的行。如果（对于给定的年、周、日和类别）您有 10 个单独的行，您希望显示哪个 MTM 值？

【参考方案1】：

尝试 SUM/AVG（取决于您的需要）PM.MTM * PG.HOURS AS ADJ_MTM 和 (PG.UNITS / (PM.MTM * PG.HOURS)) AS PERC_STANDARD，而不是按它们分组：

SELECT 
  FD.YEAR, 
  FD.WEEK, 
  PG.DT, 
  PG.CAT, 
  SUM(PG.UNITS) AS UNITS, 
  SUM(PG.HOURS) AS HOURS, 
  PM.MTM,
  SUM(PM.MTM * PG.HOURS )AS ADJ_MTM,
  SUM((PG.UNITS / (PM.MTM * PG.HOURS))) AS PERC_STANDARD,
  SUM(CASE WHEN GE.QTY IS NULL THEN 0 ELSE GE.QTY END) AS QTY,
  SUM(WH.TOTALHOURS) AS TOTALHOURS 
FROM 
  PROD_GUNS PG 
INNER JOIN PROD_MTM PM 
  ON PG.CAT = PM.CATEGORY 
  AND PM.DEPTNO = '018' 
  AND PG.DT BETWEEN PM.START_DT AND PM.END_DT 
INNER JOIN FISCAL_DATES_DAYS FD 
  ON PG.DT = FD.DT 
LEFT OUTER JOIN PROD_GUNS_ERRORS GE 
  ON PG.EID = GE.EID 
  AND PG.DT = GE.DT 
INNER JOIN WEEKLY_HOURS WH
  ON FD.WEEK = WH.DT_WEEK
  AND FD.YEAR = WH.DT_YEAR
  AND PG.EID = WH.EEXX31
GROUP BY 
  FD.YEAR, 
  FD.WEEK, 
  PG.DT, 
  PG.CAT, 
  PM.MTM 
HAVING 
  FD.YEAR = '2015' 
  AND FD.WEEK = '1' 
  AND PG.DT = '29-DEC-14' 
  AND PG.CAT = 'Picking'
ORDER BY 
  PG.DT;

【讨论】：

我也正要指出这一点。您忘记了SUM ADJ_MTM，这会将它们转换为输出中的唯一行，并防止它们汇总为一行。

@StefanYordanov 你是天才！这正是我想要的。我会在几分钟内接受你的回答。说真的，谢谢你的帮助。我花了一天半的时间试图弄清楚这一点！