At 10:53 PM -0500 7/16/01, Gilles Detillieux wrote:
>this, but in practise there are some hashing functions that are less
>likely to fail than others. I like the seed for a random number idea.
>It would be worth a try. I wonder, though, if there wouldn't be some
>other list of numbers that would have the same effect on this function.
For any set hashing function (rather than a "perfect hash" table), I
can come up with a nasty list of numbers. >:-) I can't say that the
current function is best for everything--and we might want to have
multiple hash functions (e.g. one for URLs and one for regular text
might help a lot).
I set the hash function to recognize numbers because it seemed to
save a lot on the texts I was indexing. But I'd be glad to make them
seeds for random numbers and then throw them through some sort of
mangling function--but it doesn't make sense to waste too much time
or we might as well treat them as ASCII again.
-Geoff
_______________________________________________
htdig-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/htdig-dev
