Question

使用sql server 2008

我想拿一张有1100万条记录的表格，并将它们分成每块50000块，同时保留原始表格并确保每个块包含唯一记录。如果我为第一批选择前50000条记录，我怎样才能保证我获得下一条50000等等。

Answer 1

使用排名函数，特别是row-number（）：

DECLARE
  @From  int
 ,@Thru  int

--  Example here would be the second set
SET @From = 50001
SET @Thru = 100000

SELECT <columns>
 from (select <columns>, row_number() over (order by <PrimaryKey>) Ranking
        from MyTable) xx
 where Ranking between @From and @Thru

Answer 2

尝试NTILE（n），其中n =块数。由于您需要创建“块”表，您应该知道需要多少块，而NTILE会为您正确地分割行：

DECLARE @YourTable table (RowID int, RowValue varchar(5))
INSERT @YourTable VALUES (1,'A')
INSERT @YourTable VALUES (2,'B')
INSERT @YourTable VALUES (3,'C')
INSERT @YourTable VALUES (4,'D')
INSERT @YourTable VALUES (5,'E')
INSERT @YourTable VALUES (6,'F')
INSERT @YourTable VALUES (7,'G')
INSERT @YourTable VALUES (8,'H')
INSERT @YourTable VALUES (9,'I')
INSERT @YourTable VALUES (10,'J')
INSERT @YourTable VALUES (11,'K')
INSERT @YourTable VALUES (12,'L')

DECLARE @YourTable1 table (RowID int, RowValue varchar(5))
DECLARE @YourTable2 table (RowID int, RowValue varchar(5))
DECLARE @YourTable3 table (RowID int, RowValue varchar(5))
DECLARE @YourTable4 table (RowID int, RowValue varchar(5))
DECLARE @YourTable5 table (RowID int, RowValue varchar(5))

INSERT @YourTable1
    SELECT
        RowID, RowValue
        FROM (SELECT
                  RowID, RowValue
                      ,NTILE(5) OVER(ORDER BY RowID) AS TableID
                  FROM @YourTable
             ) dt 
        WHERE dt.TableID=1


INSERT @YourTable2 
    SELECT
        RowID, RowValue
        FROM (SELECT
                  RowID, RowValue
                      ,NTILE(5) OVER(ORDER BY RowID) AS TableID
                  FROM @YourTable
             ) dt 
        WHERE dt.TableID=2

INSERT @YourTable3 
    SELECT
        RowID, RowValue
        FROM (SELECT
                  RowID, RowValue
                      ,NTILE(5) OVER(ORDER BY RowID) AS TableID
                  FROM @YourTable
             ) dt 
        WHERE dt.TableID=3


INSERT @YourTable4 
    SELECT
        RowID, RowValue
        FROM (SELECT
                  RowID, RowValue
                      ,NTILE(5) OVER(ORDER BY RowID) AS TableID
                  FROM @YourTable
             ) dt 
        WHERE dt.TableID=4


INSERT @YourTable5 
    SELECT
        RowID, RowValue
        FROM (SELECT
                  RowID, RowValue
                      ,NTILE(5) OVER(ORDER BY RowID) AS TableID
                  FROM @YourTable
             ) dt 
        WHERE dt.TableID=5

SELECT * FROM @YourTable1
SELECT * FROM @YourTable2
SELECT * FROM @YourTable3
SELECT * FROM @YourTable4
SELECT * FROM @YourTable5

输出：

RowID       RowValue
----------- --------
1           A
2           B
3           C

(3 row(s) affected)

RowID       RowValue
----------- --------
4           D
5           E
6           F

(3 row(s) affected)

RowID       RowValue
----------- --------
7           G
8           H

(2 row(s) affected)

RowID       RowValue
----------- --------
9           I
10          J

(2 row(s) affected)

RowID       RowValue
----------- --------
11          K
12          L

(2 row(s) affected)

Answer 3

我不确定你为什么要这样，但是这里 -

SELECT * FROM
( 
SELECT *, ROW_NUMBER() OVER (ORDER BY yourKey) AS Num FROM YourTable
) X WHERE x.Num BETWEEN 0 AND 49999

然后你会以编程方式增加你的0和49,999（希望，如果你要超过1100万条记录），或者手动增加。

除了sql server table int chunk

3 个答案: