SQL：使用桥 table 和 return 感兴趣列的串联列表在维度上加入事实

Question

我想将多个维度 table 加入一个事实 table。但是，我不想 returning 所有结果，而是想避免一对多关系结果。相反，我只想访问一个 table 中的一些数据，将其所有发现连接起来，并将它们 return 到一个列中，以便每个事实 1 行的预期结果仍然存在。我不希望结果中存在一对多关系。

如果我将此答案 How to concatenate text from multiple rows into a single text string in SQL Server 与其他感兴趣的列配对，我会收到一条错误消息：is invalid in the select list because it is not contained in either an aggregate function or the GROUP BY clause

-- sample values
-- bridge
dim2Key, groupKey
1,          1
2,          1
3,          1
4,          2

-- dim2
dim2Key, attributeOne
1,      'A'
2,      'B'
3,      'C'
4,      'A'

-- dim1
dim1Key, attributeTwo, attributeThree,
1,      35,                  'val1'      
2,      25,                  'val2'   
3,      45,                  'val3'   
4,      55,                  'val1'   

-- fact1
dim1Key, factvalue1, groupKey,
1,        5,            1
2,        25,           1
3,        55,           -1
4,        99,           2


-- expected values
-- fact1
dim1Key, factvalue1, groupKey, attributeTwo, attributeThree, attributeOne
1,          5,          1,          35,         'val1',         'A, B, C'
...
4,          99,         2,           55,         'val1',            'A'

Answer 1

目前还不清楚您的架构和连接应该是什么，但您似乎想要聚合 dim2 每行 fact1。

您可以在相关子查询中聚合 dim2。将其放在 APPLY 中通常更好，但您也可以直接将其放在 SELECT

中

SELECT
  fact1.dim1Key,
  fact1.factvalue1,
  fact1.groupKey,
  dim1.attributeTwo,
  dim1.attributeThree,
  dim2.attributeOne
FROM fact1
JOIN dim1 ON dim1.dim1key = fact1.dim1key
CROSS APPLY (
    SELECT attributeOne = STRING_AGG(dim2.attributeOne, ', ')
    FROM bridge b
    JOIN dim2 ON dim2.dim2key = b.dim2key
    WHERE b.groupKey = fact1.groupKey
) dim2

Answer 2

如果您无法访问 SQL 版本中的 STRING_AGG 函数，您可以使用 FOR XML PATH 来实现相同的目的。

    CREATE TABLE #bridge (dim2Key int, groupKey int)
    INSERT  #bridge (dim2Key, groupKey) 
    VALUES (1, 1)
    ,(2, 1)
    ,(3, 1)
    ,(4, 2)

    CREATE TABLE #dim2 (dim2Key int, attributeOne varchar(5))
    INSERT  #dim2 (dim2Key, attributeOne)
    VALUES (1, 'A')
    ,(2, 'B')
    ,(3, 'C')
    ,(4, 'A')

    CREATE TABLE #dim1 (dim1Key int, attributeTwo int, attributeThree varchar(5))
    INSERT #dim1 (dim1Key, attributeTwo, attributeThree)
    VALUES (1, 35,   'val1') 
    ,(2, 25,   'val2')   
    ,(3, 45,   'val3')   
    ,(4, 55,   'val1')   

    CREATE TABLE #fact1 (dim1Key int, factvalue1 int, groupKey int)
    INSERT #fact1 (dim1Key, factvalue1, groupKey)
    VALUES (1,   5,  1)
    ,(2,   25, 1)
    ,(3,   55, -1)
    ,(4,   99, 2)
    GO
    

    ;WITH pvt (groupKey, attributeOne)
    AS 
    (
        SELECT b.groupKey, d2.attributeOne 
        FROM #dim2 d2
        JOIN #bridge b
        ON d2.dim2Key = b.dim2Key
    )
    , dim2 AS
    (
        SELECT DISTINCT a.groupKey
        ,LEFT(r.attributeOne.value('text()[1]','nvarchar(max)') , LEN(r.attributeOne.value('text()[1]','nvarchar(max)'))-1) attributeOne 
        FROM pvt a
        CROSS APPLY
        (
            SELECT attributeOne + ', '
            FROM pvt r
            WHERE a.groupKey = r.groupKey
            FOR XML PATH(''), TYPE
        ) r (attributeOne)
    )



    SELECT f1.dim1Key, factvalue1, f1.groupKey, attributeTwo, attributeThree, attributeOne 
    FROM #fact1 f1
    LEFT JOIN #dim1 d1
    ON f1.dim1Key = d1.dim1Key
    LEFT JOIN dim2 d2
    ON f1.groupKey = d2.groupKey

SQL：使用桥 table 和 return 感兴趣列的串联列表在维度上加入事实

SQL: Join fact on dimension using bridge table and return concatenated list for column of interest

sql-server

azure-sql-database