如何在 T-SQL 中 CONCAT 和 GROUP BY 行

Question

这是我的桌子。 每列都表示为子查询的结果。

id | A     | B     | C     | (and many columns)
---+-------+-------+-------+----------+--------+-----+-...
 1 | 1a3b  | 1a2b  | 4a1b  |          |
 2 | 7a3b  | 3a7b  | 7a7b  |          |
 3 | 2a3b  | 1a3b  | 3a6b  |          |
 4 | 6a3b  | 6a3b  | 6a3b  |          |

我的表中有很多列，不仅有A、B和C。 结果由CONCAT方法生成。

SELECT CONCAT(A,B,C,...) FROM myTable

这就是我得到的。 不是我想要的答案。

id | A     | B     | C     | result        |
---+-------+-------+-------+---------------+
 1 | 1a3b  | NULL  | 4a1b  | 1a3b4a1b      |
 2 | 7a3b  | 3a7b  | 7a3b  | 7a3b3a7b7a3b  |
 3 | 2a3b  | 1a3b  | 3a6b  | 2a3b1a3b3a6b  |
 4 | 6a3b  | 6a3b  | 6a3b  | 6a3b6a3b6a3b  |

我找不到"group by"并获得如下结果以删除（或不 CONCAT）重复值的方法。 有没有办法做到这一点？

id | A     | B     | C     | result        |
---+-------+-------+-------+---------------+
 1 | 1a3b  | NULL  | 4a1b  | 1a3b4a1b      |
 2 | 7a3b  | 3a7b  | 7a3b  | 7a3b3a7b      |
 3 | 2a3b  | 1a3b  | 3a6b  | 2a3b1a3b3a6b  |
 4 | 6a3b  | 6a3b  | 6a3b  | 6a3b          |

Answer 1

一种基于VALUES表值构造函数和FOR XML的方法是一种选择：

桌子：

SELECT *
INTO Data
FROM (VALUES
   (1, '1a3b', NULL,   '4a1b'),
   (2, '7a3b', '3a7b', '7a3b'),
   (3, '2a3b', '1a3b', '3a6b'),
   (4, '6a3b', '6a3b', '6a3b')
) v (id, A, B, C)

陈述：

SELECT 
   *,
   result = (
      SELECT DISTINCT CONCAT('', c)
      FROM (VALUES (A), (B), (C)) v (c)
      FOR XML PATH('')   
   )
FROM Data

结果：

id A    B    C    result
-------------------------------
1  1a3b Null 4a1b 1a3b4a1b
2  7a3b 3a7b 7a3b 3a7b7a3b
3  2a3b 1a3b 3a6b 1a3b2a3b3a6b
4  6a3b 6a3b 6a3b 6a3b

Answer 2

如果我做对了

select id,A,B,C, string_agg(s, '')
from mytable 
cross apply (
     select distinct s
     from ( values(A),(B),(C)) t(s)
) t(s)
group by id,A,B,C

Answer 3

下面的查询应该工作：

DECLARE @table table(id int, a char(4), b char(4), c char(4))

insert into @table values
(2 ,'7a3b','3a7b','7a3b')

SELECT a,b,c, string_agg(t1.dval,'') WITHIN GROUP (ORDER BY a,b,c)) as conc
from @table as ot
CROSS APPLY
(
SELECT DISTINCT val from
(
VALUES
(left(a,2)),
(right(a,2)),
(left(b,2)),
(right(b,2)),
(left(c,2)),
(right(c,2))
)t(val)
) as t1(dval)
group by a,b,c