Am Montag, 20. Februar 2006 13:35 schrieb Daniel Fischer:
Cheers,
Sean
p.s. If you find any bugs, please let me know.
Re bugs:
1. printGamma [] would print an unmotivated " }", as witnessed by typeInf [] term14.
2. the case unify (ConT c) (AppT t1 t2) is missing.
and unifying a tyvar with itself fails. That probably doesn't occur in the inference-algorithm, but still...
3. too many shadowed bindings, this is always dangerous, I believe
4. I'm not sure, the datatypes are appropriate; as far as I know, expressions have a type and not a kind, which is what the use of the same Var type for Type and Exp entails.
and that led to an error: in generalise, we are interested in the free constructor-variables in the environment, not the term-variables, hence -- Free variables in ... -- ... schemes fv_scheme :: Scheme -> [Var] fv_scheme (Scheme vs ps ty) = nub (fv_preds ps ++ fv ty) \\ vs -- ... environments fv_gamma :: Gamma -> [Var] fv_gamma gamma = nub (concatMap (fv_scheme . snd) gamma) and not fv_gamma gamma = nub (map fst gamma)
I have only just glimpsed at Jones' paper, so I don't yet see, what this type inference algorithm (quite nice, btw) has to do with constructor classes. If I still don't after reading it, I'll come back to ask.
I still don't see clearly. So you've implemented the type inference algorithm from Jones' paper, good. But is there any significance or gain, apart from it being a nice and interesting exercise? Cheers, Daniel -- "In My Egotistical Opinion, most people's C programs should be indented six feet downward and covered with dirt." -- Blair P. Houghton