之前曾有一篇POST是關于用CTE實現(xiàn)Split，這種方法已經比傳統(tǒng)的方法高效了。今天我們就這個方法與CLR實現(xiàn)的Split做比較。在CLR實現(xiàn)Split函數(shù)的確很簡單，dotnet framework本身就有這個function了。
我們新建一個database project，然后建立一個UserDefinedFunctions,Code像這樣:
代碼如下:
1: ///
///
/// The STR.
/// The delimiter.
///
/// 1/8/2010 2:41 PM author: v-pliu
[SqlFunction(Name = “CLR_Split”,
FillRowMethodName = “FillRow”,
TableDefinition = “id nvarchar(10)”)]
public static IEnumerable SqlArray(SqlString str, SqlChars delimiter)
{
if (delimiter.Length == 0)
return new string[1] { str.Value };
return str.Value.Split(delimiter[0]);
}
///
///
/// The row.
/// The STR.
/// 1/8/2010 2:41 PM author: v-pliu
public static void FillRow(object row, out SqlString str)
{
str = new SqlString((string)row);
}
然后Bulid,Deploy一切OK后，在SSMS中執(zhí)行以下測試T-sql:
代碼如下:
DECLARE @array VARCHAR(max)
SET @array = ‘39,15,93,68,64,43,90,58,39,9,26,26,89,47,91,57,98,16,55,9,63,29,69,16,41,76,34,60,68,64,61,53,32,30,11,72,57,63,36,43,22,14,60,38,24,5,66,26,26,21,22,99,55,18,7,10,46,76,27,88,9,29,89,75,48,72,94,59,35,19,0,35,79,11,87,49,68,30,91,35,9,7,34,47,41,61,98,13,22,1,26,80,35,48,34,92,24,85,90,51’ SELECT id FROM dbo.CLR_Split(@array,’,’)
我們來看它的Client Statistic:
接著我們執(zhí)行測試T-sql使用相同的array:
代碼如下:
DECLARE @array VARCHAR(max)
SET @array = ‘39,15,93,68,64,43,90,58,39,9,26,26,89,47,91,57,98,16,55,9,63,29,69,16,41,76,34,60,68,64,61,53,32,30,11,72,57,63,36,43,22,14,60,38,24,5,66,26,26,21,22,99,55,18,7,10,46,76,27,88,9,29,89,75,48,72,94,59,35,19,0,35,79,11,87,49,68,30,91,35,9,7,34,47,41,61,98,13,22,1,26,80,35,48,34,92,24,85,90,51’
SELECT item FROM strToTable(@array,’,’)
CTE實現(xiàn)的Split function的Client statistic:
通過對比，你可以發(fā)現(xiàn)CLR的performance略高于CTE方式，原因在于CLR方式有Cache功能，并且把一個復雜的運算放到程序里比DataBase里更加高效。
您還可以參考：
Author:Petter Liu
