Is one hash enough? about bloom HOT 5 CLOSED

Bruce, I think Will explains why one is enough here: http://willwhim.wpengine.com/2011/09/03/producing-n-hash-functions-by-hashing-only-once/

from bloom.

Cheers land. This paper referenced from the blog post helped, although I'll admit to still wading through it http://bit.ly/rgYdK3

One 64bit is equivalent to two 32bit hashes is probably the thing I had most trouble with, but I'm getting there. It's no worse than two different hashes with the same unevenness of distribution and has the benefit that you can choose the best algorithm and not have to deal with an inferior one.

Appreciate the help. Also found http://programmers.stackexchange.com/questions/49550/which-hashing-algorithm-is-best-for-uniqueness-and-speed in my travels which was illuminating.

from bloom.

Thanks, Bruce.

I’ll give it all some thought.

-- 
Will

On November 21, 2014 at 8:00:12 AM, Bruce Fitzsimons ([email protected]) wrote:

Cheers land. This paper referenced from the blog post helped, although I'll admit to still wading through it http://bit.ly/rgYdK3

One 64bit is equivalent to two 32bit hashes is probably the thing I had most trouble with, but I'm getting there. It's no worse than two different hashes with the same unevenness of distribution and has the benefit that you can choose the best algorithm and not have to deal with an inferior one.

Appreciate the help. Also found http://programmers.stackexchange.com/questions/49550/which-hashing-algorithm-is-best-for-uniqueness-and-speed in my travels which was illuminating.

—
Reply to this email directly or view it on GitHub.

from bloom.

Will, I'm by no means an expert on hash functions, but my understanding is that FNV is not a good choice for bloom filters because it fails to properly mix all the bits of the input. This page explains why http://bretmulvey.com/hash/6.html (compare with Jenkins Hash results: http://bretmulvey.com/hash/7.html)

from bloom.

I've now done some experimenting, and I'm satisfied that the FNV hash is working well enough. If you have some specific data showing that's not the case, I'd love to see it, but I've got to put it aside for now.

from bloom.