[bitc-dev] DISCUSS: quasi-object syntax

[Swaroop Sridhar]

Jonathan S. Shapiro wrote:
> BitC is both object-disoriented and dysfunctional :-). A fairly simple
> hack would restore much of the surface idiom of object-oriented
> programming, and I would like feedback on whether it should be
> considered.
> 
> Syntax hack:
> 
>   Given an application of the form
> 
>      (a.x [args])
> 
>   we first attempt to resolve 'x' as a field name within a. If no
>   such field name exists, we rewrite the application as
> 
>     (x a [args])
> 
>   and see if we can resolve that.
> 
> This is nothing more or less than a disgusting hack, but I sort of
> suspect that it would work most of the time.
> 
> Alternative: instead of overloading "." (which is what makes this
> horrible) we could use some other form of punctuation, in which case
> this is pure sugar.

I would definitely opt for picking some other punctuation in this case.
The '.' operator is already quire overloaded:
1) Select fields from structures (including constructors)
2) Select constructor name from a type name (ex: list.cons)
3) Select bindings from interfaces

The programmer might also get unexpected error messages due to
over-overloading.

Swaroop.