A fast mostly collision free hash ?

[Pavel Sanda]

On Wed, Nov 09, 2022 at 12:29:49PM +0100, Jean-Marc Lasgouttes wrote:
> Le 08/11/2022 ?? 20:44, Pavel Sanda a écrit :
> >I do not follow what's your problem with QHash? Hash tables are designed too have
> >collisions from time to time.
> 
> My problem here is that I have a cache
>   [string, screen width] -> [points where to break the string]
> 
> In the example file submitted by Scott in #12598, the string is 500kB long,
> and therefore, this will need to be stored in the cache.
> 
> In general, I will need to store in the cache an amount of data ~equal to
> the document size, and I think it is not really necessary.
> 
> This is why I am looking for a fast and almost-collision-free hash to store
> in the key.
> 
> Is it clearer now ?

I think I understand what you need. What I do not understand why you consider
QHash not good enough for your purposes. "Almost collision-free" is another
name to get amortized complexity of O(1) for insertion/lookup and QHash
is supposed to have it, no?

Pavel