[Haskell-cafe] Capitalization and associated type families

Brian Hulley brianh at metamilk.com
Tue Aug 5 15:06:30 EDT 2008


Jonathan Cast wrote:
> 
> It's weird.  ML-derived languages are functional languages at the value
> level (and have regular lambda-bound variables in consequence), but are
> logical languages at the type level (and have logical variables in
> consequence).  So ordinary lambda-bound variables, like in ML-style
> functors (parametrized modules) act more like type constants than like
> type variables.

Thanks - the above seems to be a promising avenue of thought to arrive 
at clarity regarding case distinctions at the type level, and I can see 
here the basis for a good argument of why there would actually be no 
inconsistency regarding the use of capitals for module type members and 
lowercase for the class parameters. Perhaps this is also the root of the 
difference between associated type synonyms and class params.

Ie at the type level, uppercase == functional variable (aka "constant") 
and lowercase == logic variable...

The type level now seems to me to be quite significantly more 
complicated than the value level due to the mixing of functional and 
logic programming, not to mention that at the type level "variables" in 
an outer scope become "constants" in an inner scope as far as pattern 
matching is concerned whereas for value patterns variables are always 
fresh and never scoped (over other patterns).

Therefore I've decided to deal with the issue of case at each level 
separately since it seems immediately clear that the case distinction at 
the value level, as in Haskell/OCaml/Moby, is obviously good due to the 
"foolproof" nature of value patterns you pointed out, whereas at the 
type level it may also be good but I'll need a few more months to think 
about it... ;-)

Cheers,
Brian.



More information about the Haskell-Cafe mailing list