Is elsa underspecified? about frozen

[serge-sans-paille]

As of 1) I see this as an internal detail and I'm not sure the user should be able to control that. What makes you think it should?

The hashes obtained from elsa are used to index an array, so your move from uint64_t to size_t looks good to me. I don't see why you would need to specialize elsa. Just providing a custom hasher that permforms multiply shift would be enough, right?

I have this strange feeling I'm missing your point...

[cbeck88]

Hi, sorry that I took a long time to respond, I've been distracted. The short version is, it isn't possible to implement multiply shift hashing correctly unless you know how many bits you are supposed to output. You do the multiply, and then the high order bits have the best statistical properties, so you shift them down to produce the result. If you don't know m at the time you do the hashing, you can't do the shift part. We could just do multiply hashing but I expect it wouldn't work very well. We could try to swap all the bits from highest to lowest before we return, but that's kind of janky and I would expect it to hurt performance significantly.

I can write again later about other ideas. I think the interface needs to develop somewhat if we want people to be able to get the best results.