Re: [htdig-dev] Questions about Dictionary::hashCode

[Geoff H. <ghu...@ws...>]

>   I'm wondering if we couldn't add a String to the Dictionary class and
> use that instead of doing a malloc/strcpy everytime.. this function is
> called jillions of times.

That's probably a good idea.

>   I'm also curious as to why not use  Knuth's golden ratio hash 
> function,
> it's a well studied and known-good hash.
>
>   I can make the change and test it.. I'm just curious about the 
> rational
> for both the origional hash function and the change.

No one really cares *what* this hash function is. I certainly don't 
care if it's "well studied" if it works. The "original" hash you point 
to was one I derived in the 3.1 betas that seemed to show the fewest 
bad hashes on as many test sets I could find. One catch to this hash 
function is that it sees a lot of URLs, which means that "classic" hash 
functions don't always work well.

The change did seem to show performance improvements for me, though 
I'll admit I didn't count the number of bad hashes.

I've seen lots of "well it works" hash functions used in other 
situations, which is what I'd say about any replacements, Knuth or not.

-Geoff