Optimize access to certain data structure (string, vector, buffer, ..?)

[aaptel]

• Vector access/indexing requires a lot of boxing/unboxing
• We could provide read-only access to the internal representation of strings because it's a super set of UTF8 (mostly utf8).

[aaptel]

I've added the vec_size, vec_get, vec_set to the API. It's using ASIZE/AREF/ASET which only works on actual vectors.

Should we support vector access for string or boolean vectors like Faref() does?

[tromey]

I don't have any immediate need for boolean vectors. They aren't used very much in elisp, I think.

I think for strings it's perhaps more useful to provide read-only access to the entire string body at once, not just char-by-char.

So I would say "no" to both for now.

[phst]

We should also strive to keep the module interface as small as possible. I'd say that new functions should only be added if a real-world benchmark shows that they significantly improve performance.

[aaptel]

Maybe its time we work on a libemacs module/helper for non-core but useful stuff.