unbound functions with this as a default parameter value

[michaelficarra]

This may be a duplicate, but I can't find it:

$ coffee -bep 'do (a=@) => a'
var _this = this;

(function(a) {
  return a;
})(this);
$ coffee -bep 'do (a=@) -> a'
(function(a) {
  return a;
})(this);
$ coffee -bep '((a) => a)(@)'
var _this = this;

(function(a) {
  return a;
})(this);
$ coffee -bep '((a) -> a)(@)'
(function(a) {
  return a;
})(this);
$ coffee -bep '((a=@) => a)()'
var _this = this;

(function(a) {
  if (a == null) {
    a = _this;
  }
  return a;
})();
$ coffee -bep '((a=@) -> a)()'
(function(a) {
  if (a == null) {
    a = this;
  }
  return a;
})();
$ 

Notice that the last compilation has different behaviour from the rest, though one would expect it to have equivalent behaviour. I suspect there's similar problems with arguments, regardless of boundedness.

[michaelficarra]

Now that I think about it, this could actually be a desirable output, though syntactically confusing (this being outside the function body). I can see how someone might want a parameter to default to the context of the function.

The problem with arguments remains regardless.

[satyr]

Due to the specialcasing, do -> (or do [] = -> etc.) and do (->) (or do [] = [] = -> etc.) are different wrt the scope/type of default arguments. One of the reasons Coco introduced let in place of do ->.

[satyr]

Besides, do (a) -> and do ((a) ->) is different by definition since #960. This behavior is more like feature than bug.

[vendethiel]

Seems expected to me

[xixixao]

This is tagged bug, what is the desired output?

[michaelficarra]

I think we've come to the conclusion that default arguments are evaluated in the context of the function body, so the current output is desired. Closing as invalid.