SQL Query Group by Clause

Question

I am facing a problem with a SQL query. This is my table structure:

DECLARE @tab TABLE
             (
                   Id INT IDENTITY, 
                   Rid INT, 
                   Address1 VARCHAR(50),
                   City VARCHAR(20)
              ) 

INSERT INTO @tab VALUES (56, 'Test1', 'New York')
INSERT INTO @tab VALUES (1253, 'Test1', 'Delhi')
INSERT INTO @tab VALUES (56, 'Address5', 'Cali')
INSERT INTO @tab VALUES (1253, 'Address5', 'Delhi')

SELECT * FROM @tab

My query:

SELECT
    Address1, STUFF((SELECT ',' + CONVERT(Varchar, Id)
                     FROM @tab TR
                     WHERE TR.Rid IN (56, 1253)
                     GROUP BY Id, Rid, Address1 
                     FOR XML PATH('')), 1, 1, '') AS addid
FROM
    @tab T
WHERE 
    T.Rid IN (56,1253)
GROUP BY
    T.Address1

It is showing me all the Ids in comma wise, while I want to show the Ids Address wise like 1,2 in the first column and 3,4 in the second column, ie Ids should be group by Address1.

Thanks

Answer 1

You need to filter TR.Address1 = T.Address1 instead of TR.Rid IN (56, 1253)

SELECT
    Address1, STUFF((SELECT ',' + CONVERT(Varchar, Id)
                     FROM @tab TR
                     WHERE TR.Address1 = T.Address1 // Adjust the condition here
                     GROUP BY Id, Rid, Address1 
                     FOR XML PATH('')), 1, 1, '') AS addid
FROM
    @tab T
WHERE 
    T.Rid IN (56,1253)
GROUP BY
    T.Address1

Live demo here

Answer 2

You need a correlated subquery. I would suggest writing the query as:

SELECT Address1,
       STUFF((SELECT ',' + CONVERT(Varchar(MAX), Id)
              FROM tab TR
              WHERE TR.Address1 = T.Address1
              FOR XML PATH('')
             ), 1, 1, '') AS addid
FROM  tab T
WHERE T.Rid IN (56, 1253)
GROUP BY Address1

Notes:

• No aggregation is needed in the subquery.
• Do not use varchar without a length in SQL Server. The length varies by context and it introduces hard-to-debug errors.
• The correlation clause is tying the values back to the values in the outer query -- based on the address.
• It is unclear whether or not year really need to filter on T.Rid in the subquery. For your sample data, it is not necessary.

Here is a db<>fiddle.

In more recent versions of SQL Server, you would simply use string_agg() -- so you will be able to forget this XML hack.