[nem-en] Clarification on type-inference rules.

[Gerard Murphy]

Michal,

Five minutes after sending that last message I've realised why my previous
suggestion won't work...
 
> If this is the case, couldn't you simply put in an internal placeholder
> type
> for a local polymorphic type when compilation of the enclosing scope is
> complete?
> 
> Of course, if you had a subsequent mutative operation that would add an
> element to a polymorphically-typed array value, I presume that the process
> of type unification would pick up the type of the element and therefore
> bind
> the type variable of the array type: so you would only need to use a
> placeholder for the one specific case where the array type still contains
> an
> unbound type variable in purely functional code.
> 
> I have a feeling that I've missed something terribly obvious here - feel
> free to refer me elsewhere if this is getting painful. :-)
> 

... you would have to support upcasting a polymorphically typed array up to
System.Object or System.Array.

This then begs the question as to what the run-time type of the array should
be if you were downcasting - and how you would add an element to it!

Also, the analogy between 'array[]' and '[]' isn't exact - the former allows
mutative operations that would have to know the element type, whereas the
latter is immutable.

Doh: I think I'll go back to working through the exercises... :-)


Cheers,

Gerard