SQL Server相似比較演算法實現

概述

最近有人問到關於兩個欄位求相似度的函式，所以就寫了一篇關於相似度的函式，分別是“簡單的模糊匹配”，“順序匹配”，“一對一位置匹配”。在平時的這種函式可能會需要用到，可能業務需求不一樣，這裡只給出參照，實際情況可以相對修改。

本文所有的兩個欄位比較都是除以比較欄位本身，例如A與B比較，找出的長度除以A的長度，因為考慮如果A的長度大於B的長度，相似度會超100%，例如‘abbc’,’ab’.

如果大家想除以B的長度，只需要在語句末尾將‘SET @num=@num*1.0/LEN(@Cloumna)’修改成‘SET @num=@num*1.0/LEN(@Cloumnb)’

步驟

1.兩個欄位簡單相似

---兩個欄位簡單相似
ALTER FUNCTION DBO.FN_Resemble
    (@Cloumna NVARCHAR(MAX),
     @Cloumnb NVARCHAR(MAX)
    )
RETURNS FLOAT
AS
BEGIN
    DECLARE @num FLOAT,@len int
    SET @Cloumna=ISNULL(@Cloumna,0)
    SET @Cloumnb=ISNULL(@Cloumnb,0)
    SET @len=1
    SET @num=0
    WHILE(LEN(@Cloumna)<>0 AND LEN(@CloumnB)<>0)
    BEGIN
        WHILE(@len<=LEN(@Cloumna))
        BEGIN
            DECLARE @a NVARCHAR(4)
            SET @a=''
            SET @a=SUBSTRING(@Cloumna,@len,1)
            IF(CHARINDEX(@a,@CloumnB)>0)
            BEGIN
            SET @num=@num+1
            END
        SET  @len=@len+1   
        END
    SET @num=@num*1.0/LEN(@Cloumna)
    BREAK
    END

    RETURN @num
END

----測試程式碼
SELECT DBO.FN_Resemble('ABDC321G','ABDC123G')

2.兩個欄位順序相似

---兩個欄位順序相似
CREATE FUNCTION DBO.FN_Resemble_order
    (@Cloumna NVARCHAR(MAX),
     @Cloumnb NVARCHAR(MAX)
    )
RETURNS FLOAT
AS
BEGIN
    DECLARE @num FLOAT,@len int
    SET @Cloumna=ISNULL(@Cloumna,0)
    SET @Cloumnb=ISNULL(@Cloumnb,0)
    SET @len=1
    SET @num=0
    WHILE(LEN(@Cloumna)<>0 AND LEN(@CloumnB)<>0)
    BEGIN
    DECLARE @a NVARCHAR(4)
    DECLARE @b NVARCHAR(4)
        IF(LEN(@Cloumna)>=LEN(@CloumnB))
        BEGIN
            WHILE(@len<=LEN(@CloumnB))
            BEGIN

                SET @a=''
                SET @a=SUBSTRING(@Cloumna,@len,1)
                SET @b=''
                SET @b=SUBSTRING(@CloumnB,@len,1)
                IF(@a=@b)
                BEGIN
                SET @num=@num+1
                END
                    ELSE
                    BEGIN
                        break
                    END
            SET  @len=@len+1   
            END
        END
        ELSE IF    (LEN(@Cloumna)<LEN(@CloumnB))
        BEGIN
            WHILE(@len<=LEN(@Cloumna))
                BEGIN
                    SET @a=''
                    SET @a=SUBSTRING(@Cloumna,@len,1)
                    SET @b=''
                    SET @b=SUBSTRING(@CloumnB,@len,1)
                    IF(@a=@b)
                    BEGIN
                        SET @num=@num+1
                    END
                    ELSE
                        BEGIN
                            break
                        END
                SET  @len=@len+1   
            END

        END
    SET @num=@num*1.0/LEN(@Cloumna)
    BREAK
    END
    RETURN @num
END
go

----測試程式碼
SELECT DBO.FN_Resemble_order('ABDC456G','ABDC123G')

3.兩個欄位一對一相似

---兩個欄位一對一相似
CREATE FUNCTION DBO.FN_Resemble_onebyone
    (@Cloumna NVARCHAR(MAX),
     @Cloumnb NVARCHAR(MAX)
    )
RETURNS FLOAT
AS
BEGIN
    DECLARE @num FLOAT,@len int
    SET @Cloumna=ISNULL(@Cloumna,0)
    SET @Cloumnb=ISNULL(@Cloumnb,0)
    SET @len=1
    SET @num=0
    WHILE(LEN(@Cloumna)<>0 AND LEN(@CloumnB)<>0)
    BEGIN
    DECLARE @a NVARCHAR(4)
    DECLARE @b NVARCHAR(4)
        IF(LEN(@Cloumna)>=LEN(@CloumnB))
        BEGIN
            WHILE(@len<=LEN(@CloumnB))
            BEGIN

                SET @a=''
                SET @a=SUBSTRING(@Cloumna,@len,1)
                SET @b=''
                SET @b=SUBSTRING(@CloumnB,@len,1)
                IF(@a=@b)
                BEGIN
                SET @num=@num+1
                END
            SET  @len=@len+1   
            END
        END
        ELSE IF    (LEN(@Cloumna)<LEN(@CloumnB))
        BEGIN
            WHILE(@len<=LEN(@Cloumna))
                BEGIN
                    SET @a=''
                    SET @a=SUBSTRING(@Cloumna,@len,1)
                    SET @b=''
                    SET @b=SUBSTRING(@CloumnB,@len,1)
                    IF(@a=@b)
                    BEGIN
                        SET @num=@num+1
                    END
                SET  @len=@len+1   
            END

        END
    SET @num=@num*1.0/LEN(@Cloumna)
    BREAK
    END
    RETURN @num
END

----測試程式碼
SELECT DBO.FN_Resemble_onebyone('ABDC456G','ABDC123G')