Remove second appearence of a substring from string in SQL Server - sql-server

Question

I need to remove the second appearance of a substring from the main string, IF both substrings are next to each other. e.g.:
Jhon\Jhon\Jane\Mary\Bob needs to end Jhon\Jane\Mary\Bob
but Mary\Jane\Mary\Bob has to remain unchanged.
Can anyone can come out with a performant way to do this?
'\' is the separator of different names, so it can be use as limit of the substring to replace.
EDIT: this is to be run on a SELECT statement, so it should be a one line solution, I can't use variables.
Also, if the names are repetaed anywhere else, I have to let them there. Only remove one occurrence if both the first and the second names are the same.

Answer 1

So here is one try, but as I said, I don't think you will get a fast solution in native T-SQL.
First, if you don't already have a numbers table, create one:
SET NOCOUNT ON;
DECLARE #UpperLimit int = 4000;
;WITH n AS
(
SELECT rn = ROW_NUMBER() OVER (ORDER BY s1.[object_id])
FROM sys.all_objects AS s1
CROSS JOIN sys.all_objects AS s2
)
SELECT [Number] = rn - 1
INTO dbo.Numbers FROM n
WHERE rn <= #UpperLimit + 1;
CREATE UNIQUE CLUSTERED INDEX n ON dbo.Numbers([Number]);
Then create two functions. One that splits strings apart into a table, and then another that re-joins the results of the first function but ignores any subsequent duplicates.
CREATE FUNCTION dbo.SplitStrings
(
#List nvarchar(4000),
#Delim char(1)
)
RETURNS TABLE
AS
RETURN ( SELECT
rn = ROW_NUMBER() OVER (ORDER BY CHARINDEX(#Delim, #List + #Delim)),
[Value] = LTRIM(RTRIM(SUBSTRING(#List, [Number],
CHARINDEX(#Delim, #List + #Delim, [Number]) - [Number])))
FROM dbo.Numbers
WHERE Number <= LEN(#List)
AND SUBSTRING(#Delim + #List, [Number], 1) = #Delim
);
GO
Second function:
CREATE FUNCTION dbo.RebuildString
(
#List nvarchar(4000),
#Delim char(1)
)
RETURNS nvarchar(4000)
AS
BEGIN
RETURN ( SELECT newval = STUFF((
SELECT #Delim + x.[Value] FROM dbo.SplitStrings(#List, #Delim) AS x
LEFT OUTER JOIN dbo.SplitStrings(#List, #Delim) AS x2
ON x.rn = x2.rn + 1
WHERE (x2.rn IS NULL OR x.value <> x2.value)
ORDER BY x.rn
FOR XML PATH(''), TYPE).value(N'./text()[1]', N'nvarchar(max)'), 1, 1, N'')
);
END
GO
Now you can try it against the two samples you gave in your question:
;WITH cte(colname) AS
(
SELECT 'Jhon\Jhon\Jane\Mary\Bob'
UNION ALL SELECT 'Mary\Jane\Mary\Bob'
)
SELECT dbo.RebuildString(colname, '\')
FROM cte;
Results:
Jhon\Jane\Mary\Bob
Mary\Jane\Mary\Bob
But I strongly, strongly, strongly recommend you thoroughly test this against your typical data size before deciding to use it.

Answer 2

I decided to go for string manipulation. I thought it'd take longer to execute the query, but testing it in... ejem... production environment... ejem... I found out that it did not (much to my surprise). It ain't pretty, I know, but it's easy to mantain...
Here is a simplified version of my final query:
SELECT SOQ.PracticeId,
CASE WHEN LEFT(SOQ.myString, SOQ.SlashPos) = SUBSTRING(SOQ.myString, SOQ.SlashPos + 1, LEN(LEFT(SOQ.myString, SOQ.SlashPos)))
THEN RIGHT(SOQ.myString, LEN(SOQ.myString) - SOQ.SlashPos)
ELSE SOQ.myString
END as myString
FROM (SELECT OQ.AllFields, OQ.myString, CHARINDEX('\', OQ.myString, 0) as SlashPos
FROM MyOriginalQuery OQ) SOQ