Call-coalton-function bug/awkwardness

[Izaakwltn]

Call-coalton-function bug/awkwardness

[eliaslfox]

That is the expected behavior.

Also you don't need call-coalton-function when calling a top-level coalton function.

[eliaslfox]

Oh yeah. If your function has type class constraints then you can't call it from lisp, you need to use the coalton macro. call-coalton-function wouldn't work either in this case.

The "fix" is to define:

(coalton-toplevel
  (declare cube1-integer (Integer -> Integer))
  (define cube1-integer cube1))

Then cube1-integer can be called directly from lisp.

[Izaakwltn]

Awesome, thank you, that one works perfectly.

I'm going to add a few of these cases to the docs in the next few days, thank you for the clarity!

[stylewarning]

Regarding your last sentence, yes please do make a separate issue.

[Izaakwltn]

Sounds good, thanks for looking this over.

I'll weave some of these cases/behavior into the interop docs.

One last case/question: When calling a top-level coalton function with a parameter in lisp, is the example below the ideal/preferred/expected syntax:

(coalton-toplevel
  (define (cube1 x)
    (* x (* x x)))

  (define (print-cube1 x)
    (lisp String (x)
      (cl:format cl:nil "cubed: ~a" (coalton (cube1 (lisp Integer () x)))))))

I figured the main advantage to call-coalton-function is to avoid this syntax.

[eliaslfox]

You can just call it directly:

(define (print-cube1 x)
    (lisp String (x)
      (cl:format cl:nil "cubed: ~a" (cube1 x))))

The point of call-coalton-function is to allow calling a potentially curried function.

[Izaakwltn]

I'm afraid that solution doesn't work on my machine, though maybe it's something in my implementation.

Compiling

  (define (print-cube1 x)
    (lisp String (x)
      (cl:format cl:nil "cubed: ~a" (cube1 x))))

Returns the error

style-warning: 
    The function CUBE1 is called with one argument, but wants exactly two.
    --> PROGN DEFUN PROGN SB-IMPL::%DEFUN SB-IMPL::%DEFUN 
    --> SB-INT:NAMED-LAMBDA FUNCTION THE PROGN BLOCK THE LET VALUES 
    ==>
      1

Fixing that warning by wrapping it in coalton:

(define (print-cube1 x)
    (lisp String (x)
      (cl:format cl:nil "cubed: ~a" (coalton (cube1 x)))))

Returns

error: 
    during macroexpansion of (COALTON (CUBE1 X)). Use
    COMMON-LISP:*BREAK-ON-SIGNALS* to intercept.
    
     error: Unknown variable
      --> COALTON (/private/var/tmp/slimepQIMm3):1:16
       |
     1 |  (COALTON (CUBE1 X))
       |                  ^ unknown variable
    
    --> PROGN DEFUN PROGN SB-IMPL::%DEFUN SB-IMPL::%DEFUN 
    --> SB-INT:NAMED-LAMBDA FUNCTION THE PROGN BLOCK THE LET VALUES 
    ==>
      1
    
  style-warning: 
    The variable X is defined but never used.
    --> PROGN DEFUN PROGN SB-IMPL::%DEFUN SB-IMPL::%DEFUN 
    --> SB-INT:NAMED-LAMBDA FUNCTION THE PROGN BLOCK THE 
    ==>
      (LET ((CCF-BUG::X CCF-BUG::X-302))
        (VALUES
         (FORMAT NIL "cubed: ~a" (COALTON:COALTON (CCF-BUG::CUBE1 CCF-BUG::X)))))

Which is how I ended up at my example above, (coalton (cube1 (lisp Integer () x))).

I'd like to reiterate, I'm not really poking into this because I need this personally, or even that I think it's broken. There are enough ways to get around this with little hacks, but I just want to clarify behavior so that it can be defined clearly in documentation.