0
我一直在琢磨如何實現這個算法在處理SQL個月的實施,https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance在處理SQL的哈羅溫克勒距離算法
如何能不能做到?
我一直在琢磨如何實現這個算法在處理SQL個月的實施,https://en.wikipedia.org/wiki/Jaro%E2%80%93Winkler_distance在處理SQL的哈羅溫克勒距離算法
如何能不能做到?
今天我終於偶然發現了this堆棧溢出 - leebickmtu顯示了C#中的一個實現,最初是從Java移植過來的。我冒昧地將其移植到Transact SQL函數中,盡情享受吧!
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
現在在更多的行上試試這個,並等待.............. mots DBMS允許在C/Java /等中創建UDF是有原因的。 :-) – dnoeth
這不是真的應該在應用程序的數據層中做的事情的類型;這是一種商業/邏輯層的東西。 –
@ roryap爲什麼假定這是針對應用程序的?我受到啓發,因爲需要比較SQL查詢中的字符串,與開發中的任何類型的應用程序無關:) – Maritim
只是指出了遇到此問題的人,可能會誤導人們認爲這是一個好主意這種東西在具有數據訪問層的應用程序中。你沒有提到這件事,所以我認爲這很重要。 –