Re: [Caml-list] Safe UTF-8 string literals and pattern matching for OCaml

["Daniel Bünzli" <daniel.buenzli@erratique.ch>]

Le vendredi, 1 mai 2015 à 22:13, Gabriel Scherer a écrit :
> I wonder whether a system based on extensions [%u "..."] rather than attributes "..." [@u] could be easier to extend in the future. For example, you might want to introduce a different annotation `u16` that generates an integer array representing an utf16-encoded literal (or an abstract type of your liking, but then not in pattern position). Having an annotation change the type of the code would not be very nice.

I don't think that multiplying *representations* is a good idea and a single canonical one should be eventually chosen for this in OCaml. So personally I'm not interested in that kind of extension.

In any case desugaring the same notation to integer arrays is part of the design space I consider (though if we imagine a compiler integration it would be much less convenient). But I would never use UTF-16 for that. Using Unicode scalar values directly is less costly and leads to a conceptually correct type from a Unicode processing point of view.

Best,

Daniel