SAP Sybase ASA对表的Pivot/Unpivot的实现

针对表各列进行Pivot/Unpivot的实现，通俗来讲，就是行转列，列转行的操作。

1. 示例：

比如，

Date              Name       salary
---               ---        -----  
1/1/2012          jay          50
1/1/2012          ken          60
1/2/2012          ken          60
1/2/2012          jay          50
1/3/2012          jay          55
1/3/2012          lisa         80
....

想经过转换以后，得到下边的结果：

Date,      ken,   Jay,   Lisa 
 1/1/2012    60     50       0 
 1/2/2012    60     50       0 
 1/3/2012    0      55       80

一个支持Pivot/Unpivot语法的SQL语句，应该是下边这个样子的：

SELECT newPivot.*
    FROM (SELECT Date, Name, Salary
        FROM DBA.Salary ) AS source
    PIVOT
    ( SUM(source.Salary)
      FOR source.Name IN (‘Ken’ AS KEN, ‘Jay’ AS JAY, ‘Lisa’ AS LISA)
    ) AS newPivot

Date    KEN JAY LISA
1/1/2012    60  50  0
1/2/2012    60  50  0
1/3/2012    0   55  80

可惜的是ASA目前不支持PIVOT语法。

2. 具体实验

1) 初始化数据：

SQL语句如下：

BEGIN
DROP TABLE t1;
EXCEPTION WHEN OTHERS THEN
END;
 
CREATE TABLE t1 (
c1 VARCHAR ( 10 ) NOT NULL,
c2 VARCHAR ( 10 ) NOT NULL,
c3 INTEGER        NOT NULL,
PRIMARY KEY ( c1, c2 ) );

INSERT t1 VALUES ( 'CA', 'Q1', 1000 );
INSERT t1 VALUES ( 'CA', 'Q2', 2000 );
INSERT t1 VALUES ( 'CA', 'Q3', 9000 );
INSERT t1 VALUES ( 'CA', 'Q4', 7000 );
INSERT t1 VALUES ( 'NY', 'Q1', 4000 );
INSERT t1 VALUES ( 'NY', 'Q2', 5000 );
INSERT t1 VALUES ( 'NY', 'Q3', 1000 );
INSERT t1 VALUES ( 'NY', 'Q4', 6000 );
INSERT t1 VALUES ( 'FL', 'Q1', 9000 );
INSERT t1 VALUES ( 'FL', 'Q2', 7000 );
INSERT t1 VALUES ( 'FL', 'Q3', 2000 );
INSERT t1 VALUES ( 'FL', 'Q4', 1000 );
INSERT t1 VALUES ( 'AZ', 'Q1', 5000 );
INSERT t1 VALUES ( 'AZ', 'Q2', 5000 );
INSERT t1 VALUES ( 'AZ', 'Q3', 1000 );
INSERT t1 VALUES ( 'AZ', 'Q4', 3000 );
INSERT t1 VALUES ( 'MA', 'Q1', 2000 );
INSERT t1 VALUES ( 'MA', 'Q2', 6000 );
INSERT t1 VALUES ( 'MA', 'Q3', 5000 );
INSERT t1 VALUES ( 'MA', 'Q4', 3000 );
COMMIT;

SELECT * FROM t1 ORDER BY c1, c2;

所有的数据如下：

c1,c2,c3
'AZ','Q1',5000
'AZ','Q2',5000
'AZ','Q3',1000
'AZ','Q4',3000
'CA','Q1',1000
'CA','Q2',2000
'CA','Q3',9000
'CA','Q4',7000
'FL','Q1',9000
'FL','Q2',7000
'FL','Q3',2000
'FL','Q4',1000
'MA','Q1',2000
'MA','Q2',6000
'MA','Q3',5000
'MA','Q4',3000
'NY','Q1',4000
'NY','Q2',5000
'NY','Q3',1000
'NY','Q4',6000

用户想要看到这样的结果：

1. c2   AZ    CA    FL    MA    NY

2. Q1  5000  1000  9000  2000  4000

3. Q2  5000  2000  7000  6000  5000

4. Q3  1000  9000  2000  5000  1000

5. Q4  3000  7000  1000  3000  6000

实现的目的如下，将c1列的各午值旋转到新列里头.

BEGIN
    DECLARE @sql LONG VARCHAR;
    SET @sql = 'SELECT c2';
    FOR f_fetch
        AS c_fetch NO SCROLL CURSOR FOR
            SELECT DISTINCT t1.c1 AS @c1
            FROM t1
            ORDER BY t1.c1
            FOR READ ONLY
        DO
        SET @sql = STRING (
            @sql,
            ', SUM ( ( IF t1.c1 = ''',
            @c1,
            ''' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "',
            @c1,
            '"' );
    END FOR;
    SET @sql = STRING (
    @sql,
    ' INTO #t1 FROM t1 GROUP BY c2' );
    MESSAGE @sql TO CONSOLE;
    EXECUTE IMMEDIATE @sql;
    SELECT * FROM #t1 ORDER BY c2; -- pivot table
END;

用户如想看到下述结果：

1. c1   Q1    Q2    Q3    Q4

2. AZ  5000  5000  1000  3000

3. CA  1000  2000  9000  7000

4. FL  9000  7000  2000  1000

5. MA  2000  6000  5000  3000

6. NY  4000  5000  1000  6000

则需要将列c2各行的值进行旋转列(pivot) 操作：

BEGIN
    DECLARE @sql LONG VARCHAR;
    SET @sql = 'SELECT c1';
    FOR f_fetch
        AS c_fetch NO SCROLL CURSOR FOR
        SELECT DISTINCT t1.c2 AS @c2
        FROM t1
        ORDER BY t1.c2
        FOR READ ONLY
    DO
        SET @sql = STRING (
        @sql,
        ', SUM ( ( IF t1.c2 = ''',
        @c2,
        ''' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "',
        @c2,
        '"' );
    END FOR;
    SET @sql = STRING (
    @sql,
    ' INTO #t1 FROM t1 GROUP BY c1' );
    MESSAGE @sql TO CONSOLE;
    EXECUTE IMMEDIATE @sql;
    SELECT * FROM #t1 ORDER BY c1; -- pivot table
END;

实现的关键就是使用1或者0，进行SUM求和。 上述两个过程得到的SQL语句如下：

SELECT c2, 
SUM ( ( IF t1.c1 = 'AZ' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "AZ", 
SUM ( ( IF t1.c1 = 'CA' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "CA", 
SUM ( ( IF t1.c1 = 'FL' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "FL", 
SUM ( ( IF t1.c1 = 'MA' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "MA", 
SUM ( ( IF t1.c1 = 'NY' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "NY" 
INTO #t1 FROM t1 GROUP BY c2

SELECT c1, 
SUM ( ( IF t1.c2 = 'Q1' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "Q1", 
SUM ( ( IF t1.c2 = 'Q2' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "Q2", 
SUM ( ( IF t1.c2 = 'Q3' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "Q3", 
SUM ( ( IF t1.c2 = 'Q4' THEN 1 ELSE 0 ENDIF ) * t1.c3 ) AS "Q4" 
INTO #t1 FROM t1 GROUP BY c1

期盼着Pivot/Unpivot操作早些在ASA中实现。似乎SQLServer已经实现了此功能。