[clean-list] Clean compiler bug

[eli+@gs211.sp.cs.cmu.edu]

Martin Wierich wrote:
> Hi Eli,

Hi, thanks for your help.

> You are doing nothing illegal, it's a compiler bug. The only workaround I've
> found is to specialize your functions to certain kinds of arrays 

Hmm.  The reason I'm using Clean is its respectable array-processing
performance, which requires using unboxed.  The workaround I'm using
at the moment is simply to not use modules -- glom all my code into
one icl file.

I think it will work to define two families of functions, one boxed
and one unboxed.  Manual code replication, blah, maybe I should get
the C preprocessor involved. :-/

I wonder if this bug and its precise extent are documented anywhere.
It would be good to know what kinds of overloading will work as expected.

> (this will also speed up your code).

Oh heavens, is it doing run-time dispatch instead of instantiating at
compile time?  My reading of the "costs of overloading" section of the
manual was that this would happen only when a class was exported from
a module, but not its concrete instances.  Let me figure out the rule...

-- 
     Eli Brandt  |  eli+@cs.cmu.edu  |  http://www.cs.cmu.edu/~eli/