selecting a distinct combination of 2 columns in SQL

Question

When i run a select after a number of joins on my table I have an output of 2 columns and I want to select a distinct combination of col1 and col2 for the rowset returned.

the query that i run will be smthing like this:

select a.Col1,b.Col2 from a inner join b on b.Col4=a.Col3

now the output will be somewhat like this

now I want the output should be something like follows

its ok if I pick the second column randomly as my query output is like a million rows and I really dnt think there will be a case where I will get Col1 and Col2 output to be same even if that is the case I can edit the value..

Can you please help me with the same.. I think basically the col3 needs to be a row number i guess and then i need to selct two cols bases on a random row number.. I dont know how do i transalte this to SQL

consider the case 1a 1b 1c 1d 1e 2a 2b 2c 2d 2e now group by will give me all these results where as i want 1a and 2d or 1a and 2b. any such combination.

OK let me explain what im expecting:

with rs as(
select a.Col1,b.Col2,rownumber() as rowNumber from a inner join b on b.Col4=a.Col3)
select rs.Col1,rs.Col2 from rs where rs.rowNumber=Round( Rand() *100)

now I am not sure how do i get the rownumber or the random working correctly!!

Thanks in advance.

Martin Smith Martin Smith · Accepted Answer · 2011-03-06T13:16:21

If you simply don't care what col2 value is returned

select a.Col1,MAX(b.Col2) AS Col2
from a inner join b on b.Col4=a.Col3 
GROUP BY a.Col1

If you do want a random value you could use the approach below.

 ;WITH T
     AS (SELECT a.Col1,
                b.Col2
                ROW_NUMBER() OVER (PARTITION BY a.Col1 ORDER BY (SELECT NEWID())
                ) AS RN
         FROM   a
                INNER JOIN b
                  ON b.Col4 = a.Col3)
SELECT Col1,
       Col2
FROM   T
WHERE  RN = 1

Or alternatively use a CLR Aggregate function. This approach has the advantage that it eliminates the requirement to sort by partition, newid() an example implementation is below.

using System;
using System.Data.SqlTypes;
using System.IO;
using System.Security.Cryptography;
using Microsoft.SqlServer.Server;

[Serializable]
[SqlUserDefinedAggregate(Format.UserDefined, MaxByteSize = 8000)]
public struct Random : IBinarySerialize
{
    private MaxSoFar _maxSoFar;

    public void Init()
    {
    }

    public void Accumulate(SqlString value)
    {
        int rnd = GetRandom();
        if (!_maxSoFar.Initialised || (rnd > _maxSoFar.Rand))
            _maxSoFar = new MaxSoFar(value, rnd) {Rand = rnd, Value = value};
    }

    public void Merge(Random group)
    {
        if (_maxSoFar.Rand > group._maxSoFar.Rand)
        {
            _maxSoFar = group._maxSoFar;
        }
    }

    private static int GetRandom()
    {
        var buffer = new byte[4];

        new RNGCryptoServiceProvider().GetBytes(buffer);
        return BitConverter.ToInt32(buffer, 0);
    }

    public SqlString Terminate()
    {
        return _maxSoFar.Value;
    }

    #region Nested type: MaxSoFar

    private struct MaxSoFar
    {
        private SqlString _value;

        public MaxSoFar(SqlString value, int rand) : this()
        {
            Value = value;
            Rand = rand;
            Initialised = true;
        }

        public SqlString Value
        {
            get { return _value; }
            set
            {
                _value = value;
                IsNull = value.IsNull;
            }
        }

        public int Rand { get; set; }

        public bool Initialised { get; set; }
        public bool IsNull { get; set; }
    }

    #endregion


    #region IBinarySerialize Members

    public void Read(BinaryReader r)
    {
        _maxSoFar.Rand = r.ReadInt32();
        _maxSoFar.Initialised = r.ReadBoolean();
        _maxSoFar.IsNull = r.ReadBoolean();

        if (_maxSoFar.Initialised && !_maxSoFar.IsNull)
            _maxSoFar.Value = r.ReadString();
    }

    public void Write(BinaryWriter w)
    {
        w.Write(_maxSoFar.Rand);
        w.Write(_maxSoFar.Initialised);
        w.Write(_maxSoFar.IsNull);

        if (!_maxSoFar.IsNull)
            w.Write(_maxSoFar.Value.Value);
    }

    #endregion
}

selecting a distinct combination of 2 columns in SQL

4 Answers