Transact-SQL中Jaro-Winkler距离算法的实现
几个月来,我一直在想如何在Transact-SQL中实现此算法Transact-SQL中Jaro-Winkler距离算法的实现,sql,algorithm,tsql,jaro-winkler,Sql,Algorithm,Tsql,Jaro Winkler,几个月来,我一直在想如何在Transact-SQL中实现此算法 如何做到这一点?今天,我终于偶然发现了leebickmtu给出的堆栈溢出答案,它展示了一个C#实现,最初是从Java移植而来的。我冒昧地将它移植到了Transact-SQL函数中,祝您愉快 IF OBJECT_ID (N'dbo.InlineMax', N'FN') IS NOT NULL DROP FUNCTION dbo.InlineMax; GO CREATE FUNCTION dbo.InlineMax(@valu
如何做到这一点?今天,我终于偶然发现了leebickmtu给出的堆栈溢出答案,它展示了一个C#实现,最初是从Java移植而来的。我冒昧地将它移植到了Transact-SQL函数中,祝您愉快
IF OBJECT_ID (N'dbo.InlineMax', N'FN') IS NOT NULL
DROP FUNCTION dbo.InlineMax;
GO
CREATE FUNCTION dbo.InlineMax(@valueOne int, @valueTwo int)
RETURNS FLOAT
AS
BEGIN
IF @valueOne > @valueTwo
BEGIN
RETURN @valueOne
END
RETURN ISNULL(@valueTwo, @valueOne)
END;
GO
IF OBJECT_ID (N'dbo.InlineMin', N'FN') IS NOT NULL
DROP FUNCTION dbo.InlineMin;
GO
CREATE FUNCTION dbo.InlineMin(@valueOne int, @valueTwo int)
RETURNS FLOAT
AS
BEGIN
IF @valueOne < @valueTwo
RETURN @valueOne
RETURN @valueTwo
END;
GO
IF OBJECT_ID (N'dbo.JaroWinklerDistance', N'FN') IS NOT NULL
DROP FUNCTION dbo.JaroWinklerDistance;
GO
CREATE FUNCTION dbo.JaroWinklerDistance(@stringOne varchar(MAX), @stringTwo varchar(MAX))
RETURNS FLOAT
WITH EXECUTE AS CALLER
AS
BEGIN
DECLARE @mWeightThreshold FLOAT; SET @mWeightThreshold = 0.7;
DECLARE @mNuMChars INT; SET @mNumChars = 4;
DECLARE @lLen1 int; SET @lLen1 = LEN(@stringOne)
DECLARE @lLen2 int; SET @lLen2 = LEN(@stringTwo)
IF @lLen1 = 0
RETURN CASE WHEN @lLen2 = 0 THEN 1 ELSE 0 END
DECLARE @lSearchRange int; SET @lSearchRange = dbo.InlineMax(0,dbo.InlineMax(@lLen1, @lLen2)/2 - 1);
DECLARE @lMatched1 TABLE (position int not null, [status] bit not null)
DECLARE @lMatched2 TABLE (position int not null, [status] bit not null)
DECLARE @lNumCommon int; SET @lNumCommon = 0
DECLARE @i int; SET @i = 1; WHILE(@i <= @lLen1)
BEGIN
DECLARE @lStart int; SET @lStart = dbo.InlineMax(1, @i - @lSearchRange)
DECLARE @lEnd int; SET @lEnd = dbo.InlineMin(@i + @lSearchRange + 1, @lLen2)
DECLARE @j int; SET @j = @lStart; WHILE(@j <= @lEnd)
BEGIN
IF((SELECT [status] FROM @lMatched2 WHERE position = @j) = 1)
BEGIN
SET @j = @j + 1
CONTINUE
END
IF (SELECT SUBSTRING(@stringOne, @i, 1)) <> (SELECT SUBSTRING(@stringTwo, @j, 1))
BEGIN
SET @j = @j + 1
CONTINUE
END
INSERT INTO @lMatched1 (position, [status]) VALUES(@i, 1)
INSERT INTO @lMatched2 (position, [status]) VALUES(@j, 1)
SET @lNumCommon = @lNumCommon + 1
BREAK
END
SET @i = @i + 1
END
IF @lNumCommon = 0
BEGIN
RETURN 0.0;
END
DECLARE @lNumHalfTransposed int; SET @lNumHalfTransposed = 0
DECLARE @k INT; SET @k = 1;
DECLARE @stopLoop bit; SET @stopLoop = 0;
SET @i = 1; WHILE(@i <= @lLen1)
BEGIN
IF ((SELECT [status] FROM @lMatched1 WHERE position = @i) = 1)
BEGIN
SET @i = @i + 1
CONTINUE;
END
WHILE(@stopLoop = 0)
BEGIN
IF((SELECT [status] FROM @lMatched2 WHERE position = @k) = 0)
SET @k = @k + 1
ELSE
BREAK
IF((SELECT SUBSTRING(@stringOne, @i, 1)) <> (SELECT SUBSTRING(@stringTwo, @k, 1)))
SET @lNumHalfTransposed = @lNumHalfTransposed + 1
SET @k = @k + 1
END
SET @i = @i + 1
END
DECLARE @lNumTransposed INT; SET @lNumTransposed = @lNumHalfTransposed/2;
DECLARE @lNumCommonD FLOAT; SET @lNumCommonD = @lNumCommon
DECLARE @lWeight FLOAT; SET @lWeight = (@lNumCommonD / @lLen1 + @lNumCommonD / @lLen2 + (@lNumCommon - @lNumTransposed) / @lNumCommonD) / 3.0;
IF(@lWeight <= @mWeightThreshold)
RETURN @lWeight
DECLARE @lMax INT; SET @lMax = dbo.InlineMin(@mNumChars, dbo.InlineMin(@lLen1, @lLen2))
DECLARE @lPos INT; SET @lPos = 0
WHILE(@lPos < @lMax AND (SELECT SUBSTRING(@stringOne, @lPos, 1)) = (SELECT SUBSTRING(@stringTwo, @lPos, 1)))
BEGIN
SET @lPos = @lPos + 1
END
IF @lPos = 0
RETURN @lWeight
RETURN @lWeight + 0.1 * @lPos * (1.0 - @lWeight)
END;
GO
如果对象ID(N'dbo.InlineMax',N'FN')不为空
DROP函数dbo.InlineMax;
去
创建函数dbo.InlineMax(@valueOne int、@valueTwo int)
返回浮动
作为
开始
如果@valueOne>@valueTwo
开始
返回@valueOne
结束
返回值为空(@valueTwo,@valueOne)
结束;
去
如果对象ID(N'dbo.InlineMin',N'FN')不为空
DROP函数dbo.InlineMin;
去
创建函数dbo.InlineMin(@valueOne int,@valueTwo int)
返回浮动
作为
开始
如果@valueOne<@valueTwo
返回@valueOne
返回@valueTwo
结束;
去
如果对象ID(N'dbo.JaroWinklerDistance',N'FN')不为空
DROP函数dbo.JaroWinklerDistance;
去
创建函数dbo.JaroWinklerDistance(@stringOne-varchar(最大值),@stringTwo-varchar(最大值))
返回浮动
以EXECUTE作为调用方
作为
开始
宣布@mWeightThreshold浮动;设置@mWeightThreshold=0.7;
声明@mNuMChars INT;设置@mNumChars=4;
声明@lLen1 int;设置@lLen1=LEN(@stringOne)
声明@lLen2 int;设置@lLen2=LEN(@stringTwo)
如果@lLen1=0
当@lLen2=0时返回案例,然后1否则0结束
声明@lSearchRange int;设置@lSearchRange=dbo.InlineMax(0,dbo.InlineMax(@lLen1,@lLen2)/2-1);
声明@lMatched1表(位置int不为null,[状态]位不为null)
声明@lMatched2表(位置int不为null,[状态]位不为null)
声明@lNumCommon int;设置@lNumCommon=0
声明@i int;设置@i=1;而(@i今天,我终于偶然发现了leebickmtu给出的堆栈溢出答案,它展示了一个C#实现,最初是从Java移植的。我冒昧地将它移植到Transact-SQL函数,尽情享受吧
IF OBJECT_ID (N'dbo.InlineMax', N'FN') IS NOT NULL
DROP FUNCTION dbo.InlineMax;
GO
CREATE FUNCTION dbo.InlineMax(@valueOne int, @valueTwo int)
RETURNS FLOAT
AS
BEGIN
IF @valueOne > @valueTwo
BEGIN
RETURN @valueOne
END
RETURN ISNULL(@valueTwo, @valueOne)
END;
GO
IF OBJECT_ID (N'dbo.InlineMin', N'FN') IS NOT NULL
DROP FUNCTION dbo.InlineMin;
GO
CREATE FUNCTION dbo.InlineMin(@valueOne int, @valueTwo int)
RETURNS FLOAT
AS
BEGIN
IF @valueOne < @valueTwo
RETURN @valueOne
RETURN @valueTwo
END;
GO
IF OBJECT_ID (N'dbo.JaroWinklerDistance', N'FN') IS NOT NULL
DROP FUNCTION dbo.JaroWinklerDistance;
GO
CREATE FUNCTION dbo.JaroWinklerDistance(@stringOne varchar(MAX), @stringTwo varchar(MAX))
RETURNS FLOAT
WITH EXECUTE AS CALLER
AS
BEGIN
DECLARE @mWeightThreshold FLOAT; SET @mWeightThreshold = 0.7;
DECLARE @mNuMChars INT; SET @mNumChars = 4;
DECLARE @lLen1 int; SET @lLen1 = LEN(@stringOne)
DECLARE @lLen2 int; SET @lLen2 = LEN(@stringTwo)
IF @lLen1 = 0
RETURN CASE WHEN @lLen2 = 0 THEN 1 ELSE 0 END
DECLARE @lSearchRange int; SET @lSearchRange = dbo.InlineMax(0,dbo.InlineMax(@lLen1, @lLen2)/2 - 1);
DECLARE @lMatched1 TABLE (position int not null, [status] bit not null)
DECLARE @lMatched2 TABLE (position int not null, [status] bit not null)
DECLARE @lNumCommon int; SET @lNumCommon = 0
DECLARE @i int; SET @i = 1; WHILE(@i <= @lLen1)
BEGIN
DECLARE @lStart int; SET @lStart = dbo.InlineMax(1, @i - @lSearchRange)
DECLARE @lEnd int; SET @lEnd = dbo.InlineMin(@i + @lSearchRange + 1, @lLen2)
DECLARE @j int; SET @j = @lStart; WHILE(@j <= @lEnd)
BEGIN
IF((SELECT [status] FROM @lMatched2 WHERE position = @j) = 1)
BEGIN
SET @j = @j + 1
CONTINUE
END
IF (SELECT SUBSTRING(@stringOne, @i, 1)) <> (SELECT SUBSTRING(@stringTwo, @j, 1))
BEGIN
SET @j = @j + 1
CONTINUE
END
INSERT INTO @lMatched1 (position, [status]) VALUES(@i, 1)
INSERT INTO @lMatched2 (position, [status]) VALUES(@j, 1)
SET @lNumCommon = @lNumCommon + 1
BREAK
END
SET @i = @i + 1
END
IF @lNumCommon = 0
BEGIN
RETURN 0.0;
END
DECLARE @lNumHalfTransposed int; SET @lNumHalfTransposed = 0
DECLARE @k INT; SET @k = 1;
DECLARE @stopLoop bit; SET @stopLoop = 0;
SET @i = 1; WHILE(@i <= @lLen1)
BEGIN
IF ((SELECT [status] FROM @lMatched1 WHERE position = @i) = 1)
BEGIN
SET @i = @i + 1
CONTINUE;
END
WHILE(@stopLoop = 0)
BEGIN
IF((SELECT [status] FROM @lMatched2 WHERE position = @k) = 0)
SET @k = @k + 1
ELSE
BREAK
IF((SELECT SUBSTRING(@stringOne, @i, 1)) <> (SELECT SUBSTRING(@stringTwo, @k, 1)))
SET @lNumHalfTransposed = @lNumHalfTransposed + 1
SET @k = @k + 1
END
SET @i = @i + 1
END
DECLARE @lNumTransposed INT; SET @lNumTransposed = @lNumHalfTransposed/2;
DECLARE @lNumCommonD FLOAT; SET @lNumCommonD = @lNumCommon
DECLARE @lWeight FLOAT; SET @lWeight = (@lNumCommonD / @lLen1 + @lNumCommonD / @lLen2 + (@lNumCommon - @lNumTransposed) / @lNumCommonD) / 3.0;
IF(@lWeight <= @mWeightThreshold)
RETURN @lWeight
DECLARE @lMax INT; SET @lMax = dbo.InlineMin(@mNumChars, dbo.InlineMin(@lLen1, @lLen2))
DECLARE @lPos INT; SET @lPos = 0
WHILE(@lPos < @lMax AND (SELECT SUBSTRING(@stringOne, @lPos, 1)) = (SELECT SUBSTRING(@stringTwo, @lPos, 1)))
BEGIN
SET @lPos = @lPos + 1
END
IF @lPos = 0
RETURN @lWeight
RETURN @lWeight + 0.1 * @lPos * (1.0 - @lWeight)
END;
GO
如果对象ID(N'dbo.InlineMax',N'FN')不为空
DROP函数dbo.InlineMax;
去
创建函数dbo.InlineMax(@valueOne int、@valueTwo int)
返回浮动
作为
开始
如果@valueOne>@valueTwo
开始
返回@valueOne
结束
返回值为空(@valueTwo,@valueOne)
结束;
去
如果对象ID(N'dbo.InlineMin',N'FN')不为空
DROP函数dbo.InlineMin;
去
创建函数dbo.InlineMin(@valueOne int,@valueTwo int)
返回浮动
作为
开始
如果@valueOne<@valueTwo
返回@valueOne
返回@valueTwo
结束;
去
如果对象ID(N'dbo.JaroWinklerDistance',N'FN')不为空
DROP函数dbo.JaroWinklerDistance;
去
创建函数dbo.JaroWinklerDistance(@stringOne-varchar(最大值),@stringTwo-varchar(最大值))
返回浮动
以EXECUTE作为调用方
作为
开始
声明@mWeightThreshold FLOAT;设置@mWeightThreshold=0.7;
声明@mNuMChars INT;设置@mNuMChars=4;
声明@lLen1 int;设置@lLen1=LEN(@stringOne)
声明@lLen2 int;设置@lLen2=LEN(@stringTwo)
如果@lLen1=0
当@lLen2=0时返回案例,然后1否则0结束
声明@lSearchRange int;设置@lSearchRange=dbo.InlineMax(0,dbo.InlineMax(@lLen1,@lLen2)/2-1);
声明@lMatched1表(位置int不为null,[状态]位不为null)
声明@lMatched2表(位置int不为null,[状态]位不为null)
声明@lNumCommon int;设置@lNumCommon=0
声明@i int;设置@i=1;而(@i这并不是你应该在应用程序的数据层做的事情;这是一种业务/逻辑层的事情。@roryap为什么假设这是针对应用程序的?我之所以这样做,是因为需要比较SQL查询中的字符串,而不是与开发中的任何应用程序相关:)只是为那些遇到这种情况的人指出这一点,他们可能会被误导,认为在具有数据访问层的应用程序中执行此类操作是一个好主意。您没有对此做任何说明,因此我认为这一点很重要。这并不是您应该在应用程序的数据层中执行的类型;我认为“这是一种业务/逻辑层类型的东西。@roryap为什么假设这是针对应用程序的?我之所以这样做,是因为需要比较SQL查询中的字符串,而与开发中的任何类型的应用程序无关:)只是为那些遇到这种情况的人指出这一点,他们可能会被误导,认为在具有数据访问层的应用程序中执行此类操作是一个好主意。您没有对此做任何说明,所以我认为这一点很重要。现在在更多行上尝试这一点,然后等待…………这就是为什么mots DBMS都是这样的原因如何在C/Java/etc中创建UDF:-)现在在更多行上尝试此操作,然后等待。。。。。。。。。。。。。。mots DBMS允许在C/Java/etc中创建UDF是有原因的:-)