Creating a hashcode for use in a database (ie not using GetHashCode)

Question

I have recently been instructed in the ways of GetHashCode() and in particular \"Consumers of GetHashCode cannot rely upon it being stable over time or across appdomains\" (From

Answer 1

I would encourage you to consider what the others have said: let the database do what it's good at. Creating a hash code in order to optimize lookups is an indication that the indexes on your table aren't what they should be.

That said, if you really need a hash code:

You don't say if you want a 32-bit or 64-bit hash code. This one will create a 64-bit hash code for a string. It's reasonably collision-resistant.

public static long ComputeHashCode(string url)
{
    const ulong p = 1099511628211;

    ulong hash = 14695981039346656037;

    for (int i = 0; i < url.Length; ++i)
    {
        hash = (hash ^ url[i]) * p;
    }

    // Wang64 bit mixer
    hash = (~hash) + (hash << 21);
    hash = hash ^ (hash >> 24);
    hash = (hash + (hash << 3)) + (hash << 8);
    hash = hash ^ (hash >> 14);
    hash = (hash + (hash << 2)) + (hash << 4);
    hash = hash ^ (hash >> 28);
    hash = hash + (hash << 31);

    if (hash == (ulong)UNKNOWN_RECORD_HASH)
    {
        ++hash;
    }
    return (long)hash;
}

Note that this is a hash code and the likelihood of a collision is pretty small if you have up to a few billion records. Rule of thumb: you have a 50% chance of collision when the number of items exceeds the square root of your hash code's range. This hash code has a range of 2^64, so if you have 2^32 items, your chance of a collision is about 50%.

See http://www.informit.com/guides/content.aspx?g=dotnet&seqNum=792 and http://en.wikipedia.org/wiki/Birthday_paradox#Probability_table for more information.