Boost logo

Boost :

Subject: Re: [boost] Interest in a Bloom Filter?
From: Manuel Fiorelli (manuel.fiorelli_at_[hidden])
Date: 2009-06-17 04:40:38


2009/6/16 Milosz Marian HULBOJ <mhulboj_at_[hidden]>:
> Seems sound, but it would be good to list the typical use cases and work out
> the best generic interface for them. So far we have got following basic
> cases:
>
> Independent case:
>
> A simple case with distinct and independent hash functions that can be
>  evaluated separately:
>  h_1(x), h_2(x), h_3(x), ...
>
> Dependent cases:
>
> B case when from one hash value (i.e. 64bit) we can get multiple hash
>  values (i.e. 4*16bit):
>  h(x) -> {h_1(x), h_2(x), h_3(x), ...}
>
> C hash_combine - one hash function and the combining method
>
> D double-hashing schema
>  (http://www.eecs.harvard.edu/~kirsch/pubs/bbbf/esa06.pdf)
>  h_i(x) = h_1(x) + i*h_2(x) + optional
>  In this schema one can provide:
>  - two different hash functions (A)
>  - one hash function and use approach (B)
>  - use one hash function and do something along hash_combine (C)
>
> other, less popular (?)

Your listing of methods is really interesting....

Manuel Fiorelli


Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk