Grouping record based on a condtion

TechnologyTypeSize

XYZA200

XYZ1A200

XYZ2A300

XYZ3A300

ABC1X238

ABC2X238

PQRB320

MNOC330

I have written a query on a table whose output will look like the above. I need to know if i should store this in a record set or create a temp table to get the following fuctionality.

Now I need to concatenate the Technology based on Type and size.

As you can see in Type A we have two sizes 200 and 300.

We need to group the Technology of type A with same size together.

So the output of the procedure should be

XYZ + XYZ1

XYZ2+ XYZ3

ABC1 + ABC2etc.

We need to concatenate the Technology string with the next technology if they have the same type and size.

Can somebody please help or send any sample code.

Any help is greatly appreciated

Thanks

Swapna

CTE solution for SQL Server 2005:

With MyCTE(Size, Type, col1, col2, myNum) AS

(

SELECT a.Size, a.Type, CONVERT(varchar(50), MIN(RTRIM(a.Technology))) as col1, CONVERT(varchar(50),RTRIM((a.Technology))) as col2, 1 as myNum

FROM techTable AS a GROUP BY a.Size, a.Type, CONVERT(varchar(50),RTRIM(a.Technology))

UNION ALL

SELECT b.Size, b.Type, CONVERT(varchar(50), RTRIM(b.Technology)) as col1, CONVERT(varchar(50), (c.col2 + '+' + RTRIM(b.Technology))) as col2, c.myNum+1 as myNum

FROM techTable AS b INNER JOIN MyCTE c ON b.Size=c.Size AND b.Type= c.Type

WHERE b.Technology>c.col1

)

SELECT a.col2 As Technology_combined, a.Size, a.Type FROM MyCTE a INNER JOIN (SELECT Max(a1.myNum) as myNumMax, a1.Size, a1.Type FROM MyCTE a1

GROUP BY a1.Size, a1.Type) b on b.Size=a.Size AND b.Type= a.Type AND a.myNum= b.myNumMax

|||

you I am new to stored procedures...and working with the databse...so could you please explain the above code...I could not get much from it...Will the loop through the sample table I mentioned and return a set of concatenated Technology values....Please get back.

Thanks for your reply

Swapna

|||

and more over the data in the table is just an example...we are in no way concerned with the data in Technology Column...all we need to do is group the technology column data which have the same Type and Size

TechnologyTypeSize

XYZA200

ABCA200

ABC1A300

XYZ3A300

MNO1X238

ABC2X238

PQRB320

MNOC330

so the output should be XYZ+ABC

ABC1+XYZ3

MNO1+ABC2.... I hope I am clear now.

Please reply...Can we use cursors to do this...can someone explain how to use cursors for the above functionality

Thanks

|||

Hello:

The "techTable" would be the name of your table which holds your data.

The CTE code I posted will work in a recursive fasion.

If you are using SQL Server 2005, you can give the code a try run (remember to change the "techTable" to your table name).

|||

--CREATE TABLE MyTable(Technology VARCHAR(MAX), Type char(10), Size int)

--Enter the values suggested

--Run the following code

DECLARE @.Type CHAR(1)

DECLARE @.Size INT

DECLARE @.MyNewString CHAR(11)

DECLARE @.MyNewString2 VARCHAR(MAX)

SET @.MyNewString2 = ''

--Replace MyTable with your tablename

--Replace Technology, Type, Size with your field names

CREATE TABLE #Temp(MyNewString VARCHAR(MAX))

DECLARE c1 CURSOR FOR

SELECT mt.Type, mt.Size

FROM MyTable mt

OPEN c1

FETCH NEXT FROM c1

INTO @.Type, @.Size

WHILE @.@.FETCH_STATUS = 0

BEGIN

DECLARE c2 CURSOR FOR

SELECT Technology from MyTable Where size = @.Size and type = @.Type

OPEN c2

FETCH NEXT FROM c2

INTO @.MyNewString

WHILE @.@.FETCH_STATUS = 0

BEGIN

SET @.MyNewString2 = LTRIM(RTRIM(@.MyNewString2)) + LTRIM(RTRIM(@.MyNewString))

FETCH NEXT FROM c2

INTO @.MyNewString

END

CLOSE c2

DEALLOCATE c2

INSERT INTO #Temp(MyNewString) VALUES(@.MyNewString2)

SET @.MyNewString2 = ''

FETCH NEXT FROM c1

INTO @.Type, @.Size

END

CLOSE c1

DEALLOCATE c1

SELECT * from #Temp

GROUP BY MyNewString

DROP TABLE #temp

|||

limno wrote:

CTE solution for SQL Server 2005:

With MyCTE(Size, Type, col1, col2, myNum) AS

(

SELECT a.Size, a.Type, CONVERT(varchar(50), MIN(RTRIM(a.Technology))) as col1, CONVERT(varchar(50),RTRIM((a.Technology))) as col2, 1 as myNum

FROM techTable AS a GROUP BY a.Size, a.Type, CONVERT(varchar(50),RTRIM(a.Technology))

UNION ALL

SELECT b.Size, b.Type, CONVERT(varchar(50), RTRIM(b.Technology)) as col1, CONVERT(varchar(50), (c.col2 + '+' + RTRIM(b.Technology))) as col2, c.myNum+1 as myNum

FROM techTable AS b INNER JOIN MyCTE c ON b.Size=c.Size AND b.Type= c.Type

WHERE b.Technology>c.col1

)

SELECT a.col2 As Technology_combined, a.Size, a.Type FROM MyCTE a INNER JOIN (SELECT Max(a1.myNum) as myNumMax, a1.Size, a1.Type FROM MyCTE a1

GROUP BY a1.Size, a1.Type) b on b.Size=a.Size AND b.Type= a.Type AND a.myNum= b.myNumMax

This code is equivalent to my nested cursor approach and works, but I agree is a tad bit confusing...but nice work all the same.|||

You don't need to use cursors to get the results. Using cursors is often inefficient and consumes more resources than necessary. Very few problems require cursor based solutions and if you don't know how to use cursors that is actually good. :-) You can learn the basics of SQL to begin with than cursors.

If you are using SQL Server 2005 you can use below approach which will be faster than CTE and slightly simpler.

select t2.Type

, t2.Size

, max(case t2.seq when 1 then t1.Technology end)

+ max(case t2.seq when 2 then '+' + t2.Technology else '' end) as Technology

from (

select t1.Type, t1.Technology, t1.Size

, ROW_NUMBER() OVER(partition by t1.Type, t1.Size order by t1.Technology) as seq

from tbl as t1

) as t2

group by t2.Type, t2.Size;

You can use similar logic in older versions of SQL Server also since they don't have the ROW_NUMBER() function.

Below working query uses pubs authors table and you can do the same based on your table schema.

select a2.city, a2.state
, max(case a2.seq when 1 then a2.au_id else '' end)
+ max(case a2.seq when 2 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 3 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 4 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 5 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 6 then ', ' + a2.au_id else '' end) as au_ids
from (
select a1.city, a1.state, a1.au_id, row_number() over(partition by a1.city, a1.state order by a1.au_id) as seq
from authors as a1
) as a2
group by a2.city, a2.state
order by a2.state, a2.city;

|||

Umachandar Jayachandran - MS wrote:

You don't need to use cursors to get the results. Using cursors is often inefficient and consumes more resources than necessary. Very few problems require cursor based solutions and if you don't know how to use cursors that is actually good. :-) You can learn the basics of SQL to begin with than cursors.

If you are using SQL Server 2005 you can use below approach which will be faster than CTE and slightly simpler.

select t2.Type

, t2.Size

, min(case t2.seq when 1 then t1.Technology end)

+ min(case t2.seq when 2 then '+' + t2.Technology else '' end) as Technology

from (

select t1.Type, t1.Technology, t1.Size

, ROW_NUMBER() OVER(partition by t1.Type, t1.Size order by t1.Technology) as seq

from tbl as t1

) as t2

group by t2.Type, t2.Size;

You can use similar logic in older versions of SQL Server also since they don't have the ROW_NUMBER() function.

Not knowing cursors is a good thing? Can we go a step further with your logic and say not knowing SQL is a good thing? Use ADO?

...and could you post some working code. I'm interested in this approach but getting errors.

Thanks,

Adamus

|||

Umachandar Jayachandran - MS wrote:

You don't need to use cursors to get the results. Using cursors is often inefficient and consumes more resources than necessary. Very few problems require cursor based solutions and if you don't know how to use cursors that is actually good. :-) You can learn the basics of SQL to begin with than cursors.

If you are using SQL Server 2005 you can use below approach which will be faster than CTE and slightly simpler.

select t2.Type

, t2.Size

, min(case t2.seq when 1 then t1.Technology end)

+ min(case t2.seq when 2 then '+' + t2.Technology else '' end) as Technology

from (

select t1.Type, t1.Technology, t1.Size

, ROW_NUMBER() OVER(partition by t1.Type, t1.Size order by t1.Technology) as seq

from tbl as t1

) as t2

group by t2.Type, t2.Size;

You can use similar logic in older versions of SQL Server also since they don't have the ROW_NUMBER() function.

I unmarked this as the answer because the poster requested a cursor approach.|||

Not using procedural logic when dealing with SQL is a good thing. Yes, you can use ADO/client-side code to do this but it will be very slow and inefficient. If you have a table that contains say millions of rows you will be moving those rows from client to server for each user and performing the logic on the client side. Moreover, you have to implement lot of specific logic on the client side whereas the SQL language has built-in functionality / primitives to solve complex problems easily.

Anyway, here is a query that uses pubs authors table:

select a2.city, a2.state
, max(case a2.seq when 1 then a2.au_id else '' end)
+ max(case a2.seq when 2 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 3 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 4 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 5 then ', ' + a2.au_id else '' end)
+ max(case a2.seq when 6 then ', ' + a2.au_id else '' end) as au_ids
from (
select a1.city, a1.state, a1.au_id, row_number() over(partition by a1.city, a1.state order by a1.au_id) as seq
from authors as a1
) as a2
group by a2.city, a2.state
order by a2.state, a2.city;

The query produces a comma-separated list of author ids for each state and city combination similar to the problem.