[Haskell-cafe] ambiguous constraint errors

[Daniil Elovkov]

Hello

Evan Laforge wrote:
> I have two related questions:
> 
> #1
> 
> I'm getting some annoying type errors that I don't fully understand,
> and wind up having to do a workaround that I don't totally like.
> Here's a simplified version of my situation:
> 
> data Ambi m = Ambi {
>     ambi_monad :: m Int
>     , ambi_int :: Int
>     }
> 
> some_ambi :: Monad m => Ambi m
> some_ambi = Ambi (return 5) 10
> 
> ambi_table :: Monad m => [(String, Ambi m)]
> ambi_table = [("default", some_ambi)]
> .
> get_int :: String -> Maybe Int
> get_int sym = fmap ambi_int (lookup sym ambi_table)
> 
> -----------
> 
> get_int produces:
>     Ambiguous type variable `m' in the constraint:
>       `Monad m' arising from a use of `ambi_table' at ambi.hs:13:40-49
> 
> So I guess this means I'm not telling it which 'm', so it doesn't know
> how to resolve the 'return'... but the thing is, I'm not even using
> that value, so it doesn't matter what it resolves to.  So it works if
> I pick some random monad:
> 
> get_int sym = fmap ambi_int (lookup sym ambi_table :: Maybe (Ambi Maybe))

Of you and the type system you're the only one who knows that that value 
is not used. The type system doesn't use (all) the rules you have in 
your mind. It follows more simple ones.

You judge by values, not only types here. That is, you look at the value 
of ambi_int and see that it's just 10 in your (value again) some_ambi. 
You see that it's not

ambi_int = (some_return_from_monad ambi_monad) * 3

(If there were function returning from a monad)

In this case you wouldn't complain because the compiler definitely would 
have to know what monad it is.

Haskell type system doesn't look that far to distinguish those 2 cases. 
It doesn't deal with values (well, in a sense :).

Also compare with this

x :: Int
x = "Five"

main = putStrLn "Hello"

This program doesn't use x, so the type error would definitely not 
bother us at run-time. But it's nevertheless not ignored.



> 
> Note that I can't leave it as 'Monad m => Ambi m' because I still get
> an ambiguous type variable complaint.
> 
> I'm a little disconcerted by having to pick some random dummy monad.
> Even worse, everything this type touches starts requiring explicit
> type declarations everywhere.  Is there some easier way to do this?

I tried to fiddle with forall, but it seems a lot more simple to say 
Identity. It will be entirely local to get_int function. Maybe it's not 
so bad...

> #2
> 
> This is somewhat related to another issue I've been having, which is
> that I have some kind of complicated type, e.g. '(SomeMonad some,
> Monad m) => some (SomethingM m Status)' that I use in a lot of places.
>  It would be a lot less typing and easier to modify later if I wrote a
> type alias:
> 
> type Command = (Monad some, Monad m) => some (State.StateT () m Status)
> 
> but of course, this isn't allowed, since the type variables don't
> appear on the lhs, and if I put a context there, it's a syntax error.
> While I can write it with data:
> 
> data (Monad some, Monad m) => Command some m = Command (some
> (State.StateT () m Status))
> 
> I've been told this doesn't mean what I expect it to, which is that
> the context constraints propagate up to and unify with the containing
> type (out of curiosity, since it's accepted, what *does* this do?  I
> think I read it somewhere once, but now I forget and can't find it).
> And sure enough, using this type doesn't make my type declarations
> have the right contexts.


Well it means that you can't call any data constructor of this type with 
arguments not satisfying those constraints. Effectively it means that 
you won't ever have a value of type (Command some m) in your program 
where the pair (some,m) doesn't satisfy them.

However, the type system won't leverage that fact. And when you use a 
value of type Command some m somewhere you have to repeat the constraints.

afaik it is officially considered a Haskell mis-feature.

Am I wrong or it can be fixed by a compiler option (ghc)?

Operationally, if I get it right, it has to do with (not) attaching 
dictionaries to data constructors. If a dictionary was attached at the 
stage of constructing a Command, it could be easily re-used anywhere. If 
it's not attached you have to pass it later.

> 
> So the first problem means that I have to declare types in various
> inconvenient places, and the second one means that I have to type out
> all the various class constraints (I can still alias away the
> non-polymorphic bits), and all my type declarations start looking much
> more complicated than they are.
> 
> The "solution" I've been using for some of this is just to remove the
> polymorphism, so I can write a simple alias like
> 
> type Command = SomethingM (State.StateT () Identity Status)
> 
> and now I can think of "a command" and have various functions that
> take and return Commands, without caring that it's some kind of monad
> with context constraints.  But of course, this isn't always possible
> since sometimes I need the type to remain polymorphic (i.e. while most
> of these I don't *think* will run in some other monad, some of them
> definitely get called in multiple contexts).
> 
> Is there any nicer way around this?  And what's the underlying issue
> that makes this necessary?  I can live with all the context hair
> everywhere, but it sure would be nicer to be able to define it once
> and for all in one place.


Maybe something like

class MyAlias t1 t2 ...

instance (Monad some, Monad m, ...) => MyAlias some m ...


no where clause. It would probably require some compiler option relaxing 
type class handling a bit.


> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>