Re: [q-lang-users] H/M-based typing for Q

[Albert G. <Dr....@t-...>]

Andrew Berg wrote:
> That being said, I am in favor of a faster native code compiler.  Are there optimizations that become possible with stronger type system that are impossible without one?

Unboxing. E.g., if the type checker can infer that a certain parameter
in a recursive function is always an integer then that parameter can be
unboxed before invoking the function and can be passed around as just a
plain integer. That saves you the runtime checks for the type and
extracting the integer's value from an expression cell.

> It seems to me that if all of the equations for a program are known at compile time (and as Q is defined now, I think that is the case) then it should be reasonable to infer a lot of the type information at compile time.  No?

Yes. In fact standard H/M allows you to infer *all* types in a program,
automatically, *without* any explicit typing information (except in some
rare corner cases). Of course this becomes more difficult if you want to
support ad-hoc polymorphism (like a function or constructor which takes
either an integer or a list as an argument, which isn't possible with
standard H/M, but is a no-brainer in Q).

Albert

-- 
Dr. Albert Gr"af
Dept. of Music-Informatics, University of Mainz, Germany
Email:  Dr....@t-..., ag...@mu...
WWW:    http://www.musikinformatik.uni-mainz.de/ag