Re: [Caml-list] passing object to a function fails to compile (lablgtk2 involved)

[Ivan <ivg@ieee.org>]

Thanks, Anthony!

16.07.2012, 19:29, "Anthony Tavener" <anthony.tavener@gmail.com>:

>  Are your lines calling create_model and fill_model in a different module than where the functions are defined?
>  (or called from the toplevel?) If so, it might be that the types at the module boundary are not constrained enough
>  to be compatible? (apologies if my terminology is incorrect)

Everything happens in the same file/module. It seems to me, just by some insight, that a compiler do not know about columns, stored at the model, so, when he infers his method, they remain polymorphic. I think that some constraint should help, but I'm in doubt how to spell it right. Method set (which I'm suspecting as a cause of a failure) has type, that contains type quantifier: "'a. row:Gtk.tree_iter -> column:'a column -> 'a -> unit". Right now, type quantifiers are out of my ocaml's active vocabulary. In other words I do not know, how to handle them properly. 
Moreover, I still have no insight on the variable 'b. Why do the compiler refers to some 'b, without even mentioning it in the types it inferred? It's very strange for me. 

> If these functions are in a different module than the calls to them, 
> you might need to define the module signature, using a consistent type for a "model". 
> Sometimes two types are not equal, even though they *would* 
> amount to the same underlying representation in a particular scenario. 
> So, check the inferred types of your functions... you might need to add another hint to express what you want.

Yes, It seems that refactoring this functions in a separate modules will bring additional benefits in addition to solving the problem. I'll consider looking in that direction, though the problem of constraining unusual (for me) type still remains.