85

Possible Duplicates:
Is there an Implode type function in SQL Server 2000?
How can I group fields from multiple records in T-SQL by another field?

I have a view which I'm querying that looks like this:

BuildingName    PollNumber
------------    ----------
Foo Centre      12        
Foo Centre      13
Foo Centre      14
Bar Hall        15
Bar Hall        16
Baz School      17

I need to write a query that groups BuildingNames together and displays a list of PollNumbers like this:

BuildingName    PollNumbers
------------    -----------
Foo Centre      12, 13, 14
Bar Hall        15, 16
Baz School      17

How can I do this in T-SQL? I'd rather not resort to writing a stored procedure for this, since it seems like overkill, but I'm not exactly a database person. It seems like an aggregate function like SUM() or AVG() is what I need, but I don't know if T-SQL has one. I'm using SQL Server 2005.

5
  • Yes, this question has been asked multiple times on SO. stackoverflow.com/questions/1874966/… or stackoverflow.com/questions/3121079/… Commented Feb 17, 2011 at 15:53
  • Ah, my bad. Chalk this up as a case of not using the right search keywords then. :) Voting to close. Commented Feb 17, 2011 at 15:58
  • answered many times... but watch out, not all FOR XML PATH concatenations implementations will properly handle the XML special characters (<, &, >, etc) like my sample code (below) will... Commented Feb 17, 2011 at 16:02
  • Also a duplicate of: stackoverflow.com/questions/273238/… Commented Mar 7, 2013 at 12:48
  • 5
    SqlServer 2017 now has STRING_AGG that aggregates multiple strings into one using a given separator. Commented Aug 31, 2017 at 10:42

2 Answers 2

128

For SQL Server 2017 and up use:

STRING_AGG()

set nocount on;
declare @YourTable table (RowID int, HeaderValue int, ChildValue varchar(5))
insert into @YourTable VALUES (1,1,'CCC')
insert into @YourTable VALUES (2,2,'B<&>B')
insert into @YourTable VALUES (3,2,'AAA')
insert into @YourTable VALUES (4,3,'<br>')
insert into @YourTable VALUES (5,3,'A & Z')
set nocount off

SELECT
    HeaderValue,
    STRING_AGG(ChildValue,', ') WITHIN GROUP(ORDER BY ChildValue) AS ChildValues
FROM @YourTable
GROUP BY HeaderValue

The WITHIN GROUP(ORDER BY ...) is needed if the concatenation order is important in the result. It can be omitted otherwise.

OUTPUT:

HeaderValue ChildValues
----------- -------------
1           CCC
2           AAA< B<&>B
3           <br>, A & Z

(3 rows affected)

For SQL Server 2005 and up to 2016, you need to do something like this:

--Concatenation with FOR XML and eliminating control/encoded character expansion "& < >"
set nocount on;
declare @YourTable table (RowID int, HeaderValue int, ChildValue varchar(5))
insert into @YourTable VALUES (1,1,'CCC')
insert into @YourTable VALUES (2,2,'B<&>B')
insert into @YourTable VALUES (3,2,'AAA')
insert into @YourTable VALUES (4,3,'<br>')
insert into @YourTable VALUES (5,3,'A & Z')
set nocount off
SELECT
    t1.HeaderValue
        ,STUFF(
                   (SELECT
                        ', ' + t2.ChildValue
                        FROM @YourTable t2
                        WHERE t1.HeaderValue=t2.HeaderValue
                        ORDER BY t2.ChildValue
                        FOR XML PATH(''), TYPE
                   ).value('.','varchar(max)')
                   ,1,2, ''
              ) AS ChildValues
    FROM @YourTable t1
    GROUP BY t1.HeaderValue

OUTPUT:

HeaderValue ChildValues
----------- -------------------
1           CCC
2           AAA, B<&>B
3           <br>, A & Z

(3 row(s) affected)

Also, watch out, not all FOR XML PATH concatenations will properly handle XML special characters like my above example will.

Sign up to request clarification or add additional context in comments.

10 Comments

Thanks for this - the varchar (or nvarchar) conversion is almost always overlooked in examples of this "trick"
Great post, thanks for pointing out the handling of special XML characters.
.value('(./text())[1]', 'nvarchar(max)') performs significantly better than .value('.', 'nvarchar(max)'), in my tests by 13.2x.
It looks like there might be a function called STRING_AGG in version 2017 for this?
The STUFF works but it's is horribly slow! I can't believe there's no analytical function in SQL Server 2016 and earlier to handle this.
|
36

There is no built in function in Sql Server, but it can be achieved by writing a user defined aggregate. This article mentions such a function as part of the SQL Server samples: http://msdn.microsoft.com/en-us/library/ms182741.aspx

As an example I include the code for a Concatenate aggregate. To use it, create a database project in Visual Studio, add new SqlAggregate and replace the code with the sample below. Once deployed you should find a new assembly in your database and an aggregate function Concatenate

using System;
using System.Data.SqlTypes;
using System.IO;
using System.Text;
using Microsoft.SqlServer.Server;

[Serializable]
[SqlUserDefinedAggregate(Format.UserDefined, IsInvariantToNulls = true, IsInvariantToDuplicates = false, IsInvariantToOrder = false, MaxByteSize = 8000, Name = "Concatenate")]
public class Concatenate : IBinarySerialize
{
    private StringBuilder _intermediateResult;

    internal string IntermediateResult {
        get
        {
            return _intermediateResult.ToString();
        } 
    }

    public void Init()
    {
        _intermediateResult = new StringBuilder();
    }

    public void Accumulate(SqlString value)
    {
        if (value.IsNull) return;
        _intermediateResult.Append(value.Value);
    }

    public void Merge(Concatenate other)
    {
        if (null == other)
            return;

        _intermediateResult.Append(other._intermediateResult);
    }

    public SqlString Terminate()
    {
        var output = string.Empty;

        if (_intermediateResult != null && _intermediateResult.Length > 0)
            output = _intermediateResult.ToString(0, _intermediateResult.Length - 1);

        return new SqlString(output);
    }

    public void Read(BinaryReader reader)
    {
        if (reader == null) 
            throw new ArgumentNullException("reader");

        _intermediateResult = new StringBuilder(reader.ReadString());
    }

    public void Write(BinaryWriter writer)
    {
        if (writer == null) 
            throw new ArgumentNullException("writer");

        writer.Write(_intermediateResult.ToString());
    }
}

To use it, you can simply write an aggregate query:

create table test(
  id int identity(1,1) not null
    primary key
, class tinyint not null
, name nvarchar(120) not null )

insert into test values 
(1, N'This'),
(1, N'is'),
(1, N'just'),
(1, N'a'),
(1, N'test'),
(2, N','),
(3, N'do'),
(3, N'not'),
(3, N'be'),
(3, N'alarmed'),
(3, N','),
(3, N'this'),
(3, N'is'),
(3, N'just'),
(3, N'a'),
(3, N'test')


select dbo.Concatenate(name + ' ')
from test
group by class

drop table test

The output of the query is:

-- Output
-- ===================
-- This is just a test
-- ,
-- do not be alarmed , this is just a test

I packaged up the class and the aggregate as a script which you can find here: https://gist.github.com/FilipDeVos/5b7b4addea1812067b09

13 Comments

+1 this is much more helpful for me than the chosen answer. It's exactly what I need.
@SerjSagan I added the code + an example and link to an installation script so it is clearer for people reading this answer.
With this approach, how do I control the order in which items are aggregated? Seems the order will be determined by the clustered index.
You can't, unfortunately user defined aggregates do not support order by in windowing functions. An option to force order is to use an index hint WITH (INDEX(pk_test))
This is a great way to add maintainability barriers to your product.
|

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.