It can be done w/o a cursor. For example,

IF object_id('tempdb..#temp01') IS NOT NULL
	DROP TABLE #temp01
	
SELECT *, IDENTITY(int,1,1) AS pk  INTO #temp01 FROM mytable	

--SELECT * FROM #temp01 

SELECT 
		(SELECT col1 FROM #temp01 WHERE pk=1) AS NewCol1,
		(SELECT col1 FROM #temp01 WHERE pk=2) AS NewCol2,
		(SELECT col1 FROM #temp01 WHERE pk=3) AS NewCol3
UNION ALL
SELECT 
		(SELECT col2 FROM #temp01 WHERE pk=1) AS NewCol1,
		(SELECT col2 FROM #temp01 WHERE pk=2) AS NewCol2,
		(SELECT col2 FROM #temp01 WHERE pk=3) AS NewCol3
UNION ALL
SELECT 
		(SELECT col3 FROM #temp01 WHERE pk=1) AS NewCol1,
		(SELECT col3 FROM #temp01 WHERE pk=2) AS NewCol2,
		(SELECT col3 FROM #temp01 WHERE pk=3) AS NewCol3
UNION ALL
SELECT 
		(SELECT col4 FROM #temp01 WHERE pk=1) AS NewCol1,
		(SELECT col4 FROM #temp01 WHERE pk=2) AS NewCol2,
		(SELECT col4 FROM #temp01 WHERE pk=3) AS NewCol3

This code is not optimized but should give you a starting point.

>I want to create a new table from a existing table. In the new table the row data is column data of old table. For example:
>
>1 1 1 1
>2 2 2 2
>3 3 3 3
>
>
>1 2 3
>1 2 3
>1 2 3
>1 2 3
>
>
>I think we could use cursor in SQL to do this.
>