Foreign definitions for modules Common, CodeUnits and Unsafe

[postsolar]

These come together because they are needed to be able to reproduce a minimal tests' subset run. Imports imply/depend on merging of the IrRegex PR, as well as SRFI N152 PR for purescm/purescm.

[anttih]

Small stylistic nitpick: could you use square brackets in the let form for the binding pairs?

[anttih]

With this we also arrive at the spot where we need to figure out how to deal with the fact that JS strings are UTF-16 encoded code units and Chez uses unicode code points:

> "😁a😁😁".charAt(1)
'\ude01'

> (string-ref "😁a😁😁" 1)
#\a

For example the CodeUnit module now deals with code points rather than code units. We could actually make it work with code units by turning the string to a bytevector with string->utf16 and indexing into that.

[postsolar]

With this we also arrive at the spot where we need to figure out how to deal with the fact that JS strings are UTF-16 encoded code units and Chez uses unicode code points:

Something like this?

  (define _charAt
    (lambda (just)
      (lambda (nothing)
        (lambda (i)
          (lambda (s)
            (if (< i (string-length s))
              (let
                ([cus (bvs:string->utf16 s)]
                 [v (bvs:make-bytevector 2)]
                 [ix (* i 2)]
                 [tx (make-transcoder (utf-16-codec))])
                (bvs:bytevector-u16-set!
                  v
                  0
                  (bvs:bytevector-u16-ref cus ix (bvs:native-endianness))
                  (bvs:native-endianness))
                (just (string-ref (bytevector->string v tx) 0)))
              nothing))))))

[postsolar]

Ok I think it's ready.

[postsolar]

A note to myself: for SRFI 152, potentially there's a huge performance gain in using substring primitive instead of take and especially drop.