CHECKSUM and CHECKSUM_AGG: What's the algorithm?

Question

We perform checksums of some data in sql server as follows:

declare @cs int;
select 
    @cs = CHECKSUM_AGG(CHECKSUM(someid, position))
from 
    SomeTable
where 
    userid = @userId
group by 
    userid;

This data is then shared with clients. We'd like to be able to repeat the checksum at the client end... however there doesn't seem to be any info about how the checksums in the functions above are calculated. Can anyone enlighten me?

Answer 1

On SQL Server Forum, at this page, it's stated:

The built-in CHECKSUM function in SQL Server is built on a series of 4 bit left rotational xor operations. See this post for more explanation.

Answer 2

The CHECKSUM function doesn't provide a very good quality checksum and IMO is pretty useless for most purposes. As far as I know the algorithm isn't published. If you want a check that you can reproduce yourself then use the HashBytes function and one of the standard, published algorithms such as MD5 or SHA.

Answer 3

//Quick hash sum of SQL and C # mirror Ukraine

     private Int64 HASH_ZKCRC64(byte[] Data)
    {
        Int64 Result = 0x5555555555555555;
        if (Data == null || Data.Length <= 0) return 0;
        int SizeGlobalBufer = 8000;
        int Ost = Data.Length % SizeGlobalBufer;
        int LeftLimit = (Data.Length / SizeGlobalBufer) * SizeGlobalBufer;

        for (int i = 0; i < LeftLimit; i += 64)
        {
            Result = Result
            ^ BitConverter.ToInt64(Data, i)
            ^ BitConverter.ToInt64(Data, i + 8)
            ^ BitConverter.ToInt64(Data, i + 16)
            ^ BitConverter.ToInt64(Data, i + 24)
            ^ BitConverter.ToInt64(Data, i + 32)
            ^ BitConverter.ToInt64(Data, i + 40)
            ^ BitConverter.ToInt64(Data, i + 48)
            ^ BitConverter.ToInt64(Data, i + 56);
             if ((Result & 0x0000000000000080) != 0)
             Result = Result ^ BitConverter.ToInt64(Data, i + 28); 
        }

        if (Ost > 0)
        {
           byte[] Bufer = new byte[SizeGlobalBufer];
           Array.Copy(Data, LeftLimit, Bufer, 0, Ost);
           for (int i = 0; i < SizeGlobalBufer; i += 64)
           {
               Result = Result
               ^ BitConverter.ToInt64(Bufer, i)
               ^ BitConverter.ToInt64(Bufer, i + 8)
               ^ BitConverter.ToInt64(Bufer, i + 16)
               ^ BitConverter.ToInt64(Bufer, i + 24)
               ^ BitConverter.ToInt64(Bufer, i + 32)
               ^ BitConverter.ToInt64(Bufer, i + 40)
               ^ BitConverter.ToInt64(Bufer, i + 48)
               ^ BitConverter.ToInt64(Bufer, i + 56);
               if ((Result & 0x0000000000000080)!=0)
               Result = Result ^ BitConverter.ToInt64(Bufer, i + 28); 
           }
        }

        byte[] MiniBufer = BitConverter.GetBytes(Result);
        Array.Reverse(MiniBufer);
        return BitConverter.ToInt64(MiniBufer, 0);

        #region SQL_FUNCTION
        /*  CREATE FUNCTION [dbo].[HASH_ZKCRC64] (@data as varbinary(MAX)) Returns bigint
            AS
            BEGIN
            Declare @I64 as bigint Set @I64=0x5555555555555555
            Declare @Bufer as binary(8000)
            Declare @i as int Set @i=1
            Declare @j as int 
            Declare @Len as int Set @Len=Len(@data)     

            if ((@data is null) Or (@Len<=0)) Return 0

              While @i<=@Len
              Begin
               Set @Bufer=Substring(@data,@i,8000)
               Set @j=1
                   While @j<=8000
                   Begin
                    Set @I64=@I64 
                    ^ CAST(Substring(@Bufer,@j,   8) as bigint) 
                    ^ CAST(Substring(@Bufer,@j+8, 8) as bigint) 
                    ^ CAST(Substring(@Bufer,@j+16,8) as bigint)
                    ^ CAST(Substring(@Bufer,@j+24,8) as bigint)
                    ^ CAST(Substring(@Bufer,@j+32,8) as bigint)
                    ^ CAST(Substring(@Bufer,@j+40,8) as bigint)
                    ^ CAST(Substring(@Bufer,@j+48,8) as bigint)
                    ^ CAST(Substring(@Bufer,@j+56,8) as bigint)
                    if @I64<0 Set @I64=@I64 ^ CAST(Substring(@Bufer,@j+28,8) as bigint)      
                    Set @j=@j+64    
                   End;  
               Set @i=@i+8000
              End
            Return @I64
            END
         */
        #endregion

   }