[mercury-users] Higher order stuff, type inference and ambiguity

[Ralph Becket]

I've been hacking away and the following line

ArgSpace = array(list__map(map__lookup(TypeObjsMap), TypeSigList))

led to this collection of errors:

naffplan.m:637: In clause for function `naffplan:qualify_op_spec/2':
naffplan.m:637:   in argument 1 of functor `array/1':
naffplan.m:637:   in argument 1 of functor `list:map/2':
naffplan.m:637:   in unification of argument
naffplan.m:637:   and term `map:lookup(TypeObjsMap)':
naffplan.m:637:   type error in argument(s) of functor `map:lookup/1'.
naffplan.m:637:   argument has type `((func PropName) = T)',
naffplan.m:637:   functor `map:lookup/1' has overloaded type
naffplan.m:637:   { (map:lookup((tree234:tree234(K, V)))) :: pred(K, V),
naffplan.m:637:   (map:lookup((tree234:tree234(K, V)))) :: ((func K) = V) },
naffplan.m:637:   variable `TypeObjsMap' has type `(tree234:tree234(Type,
(list:list(Obj))))'.

I've been bitten by this one a couple of times before.  The problem is
that map__lookup/3 is a predicate and map__lookup/2 is a function.  Now,
there is sufficient information later on in the clause containing the 
offending line to show that the only consistent typing results from
taking map__lookup(TypeObjsMap) as a function.

It seems to me (as a non-expert on type inference) that provided there is
only one sensible assignment of types to values then the compiler should
settle on that one as being what you intended.

Is my problem above a special case of a much harder problem or is this
an undesirable behaviour to want or is it just a hole in type inference
in the compiler?

Cheers,

Ralph

--
Ralph Becket      |      MSR Cambridge      |      rbeck at microsoft.com 
--------------------------------------------------------------------------
mercury-users mailing list
post:  mercury-users at cs.mu.oz.au
administrative address: owner-mercury-users at cs.mu.oz.au
unsubscribe: Address: mercury-users-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-users-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------