std::collate hash() still broken

[Paolo Carlini]

AWLaFramboise@aol.com wrote:

>Now.. whether 7 is the number to use or not.. i have no idea.. I've seen 5 and 6 used as well..
>but rot7 was what was here before so..

I see. I'm not an hashing expert either, but what you are saying sounds ok to me, after cross checking with a couple of other implementations too.

Personally, I would like to see cited in a comment a relevant piece of literature on "rotary hashing", where it is at least explained why __val << X implies that the following __val >> Y must have Y = digits - X.

Thanks,
Paolo.