Performance of hashes and associative arrays

[Manfred Nowak]

Raphaël Jakse wrote:

> Is compressing for performance reasons?
Hopefully. Because in the general case the distribution of the 
keys is unknown, no function used for computing the hash value 
can be guarenteed to indeed spread the hash values uniformly 
over the hash interval.

Compressing would have a negative effect on performance if more 
time for compressing and decrompessing would be needed than time 
was lost for resolving conflicting hash values. 

> is it more interesting to compress and then hash than just
> hash ?
The only purpose of compressing is to create the special case of 
close to no redundancy in the keys used for hashing. If this 
special case is already confirmed or it is known that no 
compression will deminish the computing time: just hash.

-manfred