Incorrect output for FFI definitions with additional haskell arguments

[davidsd]

The following code in JQuery.hs:

data CloneType = WithoutDataAndEvents | WithDataAndEvents | DeepWithDataAndEvents

clone :: CloneType -> JQuery -> Fay JQuery
clone WithoutDataAndEvents  = ffi "%2['clone'](false)"
clone WithDataAndEvents     = ffi "%2['clone'](true, false)"
clone DeepWithDataAndEvents = ffi "%2['clone'](true, true)"

leads to the following JS output:

var JQuery$clone = function($p1){
  return new Fay$$$(function(){
    if (Fay$$_($p1) instanceof $_JQuery$WithoutDataAndEvents) {
      return Fay$$_(JQuery$ffi)(Fay$$list("%2['clone'](false)"));
    }
    if (Fay$$_($p1) instanceof $_JQuery$WithDataAndEvents) {
      return Fay$$_(JQuery$ffi)(Fay$$list("%2['clone'](true, false)"));
    }
    if (Fay$$_($p1) instanceof $_JQuery$DeepWithDataAndEvents) {
      return Fay$$_(JQuery$ffi)(Fay$$list("%2['clone'](true, true)"));
    }
    throw ["unhandled case in clone",[$p1]];
  });
};

[chrisdone]

So ffi doesn't take any additional arguments. If you want this to work you could open an issue or submit a patch, it's totally doable, you'd just adjust the requirements of the ffi command and have it strip the first n types from the signature and embed the ffi function inside a normalish function. You could even have any ffi expression just take the type of the nearest parent type signature. Not sure I feel like writing this code presently, still recovering from illness, but it sounds like a natural and useful extension.

[bergmark]

This works if you give each ffi call its own type signature.