Sql server SQL Server:按所有列分组

我有一个有16列的表和超过100万条记录（每月都会批量添加大约20万条记录）

大概是这样的：

+----+------+------+-----+--------+----------------+
| Id | Col2 | Col3 | ... | Col 15 | Numeric Column |
+----+------+------+-----+--------+----------------+
|  1 | xxx  | xxx  | ... | xxx    |             10 |
|  2 | yyy  | yyy  | ... | yyy    |             20 |
|  3 | zzz  | zzz  | ... | zzz    |             30 |
|  4 | xxx  | xxx  | ... | xxx    |             40 |
+----+------+------+-----+--------+----------------+

大多数记录都是唯一的，但也有重复的。 我需要在这个表上创建一个视图来聚合这些重复的行，考虑到每一列。 聚合记录时，我需要显示最后一个数字列的总和。 例如，我的视图需要聚合记录1和记录4。 预期结果是（注意50（10+40）之和）：

我通过以下查询完成了此操作：

SELECT
    [Column 2],
    [Column 3],
    ...,
    [Column 15],
    SUM([Numeric Column]) AS [Sum Of Numeric Column],
    COUNT(*) AS [Amount Of Aggregated Records]
FROM dbo.MyLargeTable WITH (NOLOCK)
GROUP BY [Column 2],
    [Column 3],
    ...,
    [Column 15]

此视图将作为需要聚合数据的其他视图的基础视图。这意味着它需要非常快。我的问题很简单：知道这个视图需要为其他视图提供数据，这是正确的方法吗？有没有办法优化此查询？我可以安排索引吗

---

谢谢。

不太确定，但想法是创建2个CTE，仅在需要时计算聚合。您可以对此进行测试，并将其与您自己的查询进行比较

WITH Dups AS 
(
SELECT
    [Column 2],
    [Column 3],
    ...,
    [Column 15],
    SUM([Numeric Column]) AS [Sum Of Numeric Column],
    COUNT(*) AS [Amount Of Aggregated Records]
FROM dbo.MyLargeTable WITH (NOLOCK)
GROUP BY [Column 2],
    [Column 3],
    ...,
    [Column 15]
HAVING COUNT(*) > 1
),

REGS AS 
(
SELECT
    [Column 2],
    [Column 3],
    ...,
    [Column 15],
    [Numeric Column],
    1 AS [Amount Of Aggregated Records]
FROM dbo.MyLargeTable WITH (NOLOCK)
GROUP BY [Column 2],
    [Column 3],
    ...,
    [Column 15]
HAVING COUNT(*) = 1
)

SELECT * FROM Regs UNION ALL SELECT * FROM Dups ORDER BY SomeCol

---

据我所知，您的查询已经优化，不需要进一步优化。在您的情况下，使用CTE只会降低性能

你有什么问题？到目前为止还没有真正的问题。我只是想就如何使用GROUP BY here来聚合重复行征求意见。查询速度已经相当慢，而且考虑到将来会增加数百万行，我希望得到最好的解决方案。然后访问

WITH Dups AS 
(
SELECT
    [Column 2],
    [Column 3],
    ...,
    [Column 15],
    SUM([Numeric Column]) AS [Sum Of Numeric Column],
    COUNT(*) AS [Amount Of Aggregated Records]
FROM dbo.MyLargeTable WITH (NOLOCK)
GROUP BY [Column 2],
    [Column 3],
    ...,
    [Column 15]
HAVING COUNT(*) > 1
),

REGS AS 
(
SELECT
    [Column 2],
    [Column 3],
    ...,
    [Column 15],
    [Numeric Column],
    1 AS [Amount Of Aggregated Records]
FROM dbo.MyLargeTable WITH (NOLOCK)
GROUP BY [Column 2],
    [Column 3],
    ...,
    [Column 15]
HAVING COUNT(*) = 1
)

SELECT * FROM Regs UNION ALL SELECT * FROM Dups ORDER BY SomeCol