sql-server - 将垂直结果转换为水平模式(T-SQL)

Question

以下是示例数据:

计算日期PLResult
2014-01-02       100         
2014-01-03       200         
2014-02-03       300         
2014-02-04       400         
2014-02-27       500         

以下是预期结果(以逻辑格式):

一月                                 二月                                 
计算日期PLResult计算日期PLResult  
2014-01-02       100         2014-02-03       300          
2014-01-03       200         2014-02-04       400          
                                         2014-02-27       500          

以下是预期结果(使用 T-SQL 查询):

一月计算日期一月-PL结果二月-计算日期二月-PL结果  
2014-01-02              100                2014-02-03              300                   
2014-01-03              200                2014-02-04              400                   
                                                      2014-02-27              500                  

目标:

• 根据月份对结果进行分类。在上面的示例中，一月的结果放置在一月细目中。
• 月数可以是动态的。在上面的示例中，仅显示一月和二月，因为只有 2 个月的结果
• 结果将通过Excel显示。其实我可以查询多个查询表来聚合不同月份的结果，但如果可以通过一次查询返回所有结果，那么维护和调试会更容易。

以下是填充示例数据的脚本:

CREATE TABLE #PLResultPerDay ( CalculationDate DATETIME, PLResult DECIMAL(18,8) )
INSERT INTO #PLResultPerDay ( CalculationDate, PLResult ) VALUES ('2014-01-02' , 100 )
INSERT INTO #PLResultPerDay ( CalculationDate, PLResult ) VALUES ('2014-01-03' , 200 )
INSERT INTO #PLResultPerDay ( CalculationDate, PLResult ) VALUES ('2014-02-03' , 300 )
INSERT INTO #PLResultPerDay ( CalculationDate, PLResult ) VALUES ('2014-02-04' , 400 )

到目前为止，这是我构建查询的尝试:

SELECT 
    CalculationDate, [January], CalculationDate, [February]
FROM 
(
    SELECT CalculationDate, PLResult, DATENAME(MONTH, CalculationDate) AS [MTH]
    FROM #PLResultPerDay
) x
PIVOT
( 
    MIN(PLResult)
    FOR [MTH] IN ([January], [February])
) p

Answer 1

正如前面所说，这实际上是不可能的，您可以获得的最接近的是:

January2014CalculationDate | January2014PLResult | February2014CalculationDate | February2014PLResult
---------------------------+---------------------+-----------------------------+------------------
    2014-01-02             |       100           |       2014-02-03            |       300
    2014-01-03             |       200           |       2014-02-04            |       400
    NULL                   |       NULL          |       2014-02-27            |       500

即使这并不简单，我仍然建议在 sql 之外处理这样的格式。第一步是按月份对数据进行分区，然后对每个月中的日期进行排名:

SELECT  CalculationDate,
        PLResult,
        CalculationMonth,
        DenseRank = DENSE_RANK() OVER(PARTITION BY CalculationMonth ORDER BY CalculationDate)
FROM    (   SELECT  CalculationDate,
                    PLResult,
                    CalculationMonth = DATEADD(MONTH, DATEDIFF(MONTH, 0, CalculationDate), 0)
            FROM    #PLResultPerDay
        ) pl;

这给出:

CalculationDate PLResult    CalculationMonth    DenseRank
2014-01-02      100         2014-01-01          1
2014-01-03      200         2014-01-01          2
2014-02-03      300         2014-02-01          1
2014-02-04      400         2014-02-01          2
2014-02-27      500         2014-02-01          3

然后您可以透视此数据:

WITH Data AS
(   SELECT  CalculationDate,
            PLResult,
            CalculationMonth,
            DenseRank = DENSE_RANK() OVER(PARTITION BY CalculationMonth ORDER BY CalculationDate)
    FROM    (   SELECT  CalculationDate,
                        PLResult,
                        CalculationMonth = DATEADD(MONTH, DATEDIFF(MONTH, 0, CalculationDate), 0)
                FROM    #PLResultPerDay
            ) pl
)
SELECT  Jan2014CalcDate = MIN(CASE WHEN CalculationMonth = '20140101' THEN CalculationDate END),
        Jan2014Result = SUM(CASE WHEN CalculationMonth = '20140101' THEN PLResult END),
        Feb2014CalcDate = MIN(CASE WHEN CalculationMonth = '20140201' THEN CalculationDate END),
        Feb2014Result = SUM(CASE WHEN CalculationMonth = '20140201' THEN PLResult END)
FROM    Data
GROUP BY DenseRank
ORDER BY DenseRank;

这给出:

Jan2014CalcDate Jan2014Result   Feb2014CalcDate Feb2014Result
2014-01-02      100             2014-02-03      300
2014-01-03      200             2014-02-04      400
NULL            NULL            2014-02-27      500

由于您的月份数是动态的，因此您需要动态构建上述语句并使用 SP_EXECUTESQL 来运行它:

DECLARE @SQL NVARCHAR(MAX) = '';

WITH Months AS
(   SELECT  M,
            ColName = DATENAME(MONTH, M) + DATENAME(YEAR, M),
            CharFormat = CONVERT(VARCHAR(8), M, 112)
    FROM    (   SELECT  DISTINCT M = DATEADD(MONTH, DATEDIFF(MONTH, 0, CalculationDate), 0)
                FROM    #PLResultPerDay
            ) m
)
SELECT  @SQL = 'WITH Data AS
                (   SELECT  CalculationDate,
                            PLResult,
                            CalculationMonth,
                            DenseRank = DENSE_RANK() OVER(PARTITION BY CalculationMonth ORDER BY CalculationDate)
                    FROM    (   SELECT  CalculationDate,
                                        PLResult,
                                        CalculationMonth = DATEADD(MONTH, DATEDIFF(MONTH, 0, CalculationDate), 0)
                                FROM    #PLResultPerDay
                            ) pl
                )
                SELECT  ' + 
                STUFF(( SELECT  ', ' + ColName + 'CalculationDate = MIN(CASE WHEN CalculationMonth = ''' + CharFormat + ''' THEN CalculationDate END), ' + 
                                ColName + 'PLResult = SUM(CASE WHEN CalculationMonth = ''' + CharFormat + ''' THEN PLResult END)'
                        FROM    Months
                        ORDER BY M
                        FOR XML PATH(''), TYPE
                    ).value('.', 'NVARCHAR(MAX)'), 1, 2, '') + 
                'FROM   Data
                GROUP BY DenseRank
                ORDER BY DenseRank;';

EXECUTE SP_EXECUTESQL @SQL;

<强> Example on SQL Fiddle

请注意，我仍然建议不要使用这种技术，并认为 SQL 应该留给存储/检索数据，以及用于格式化数据的表示层