[Haskell-beginners] Ambiguous MonadIO and Monad

[Ken KAWAMOTO]

Thanks Daniel.

Adding a type signature to response did resolve the issue.
The monomorphism restriction
(http://www.haskell.org/haskellwiki/Monomorphism_restriction)
is still a bit hard for me, but I'll definitely add type signatures to
all top-level defs to avoid this kind of issues.

-- Ken

On Wed, Mar 7, 2012 at 9:55 AM, Daniel Fischer
<daniel.is.fischer at googlemail.com> wrote:
> On Wednesday 07 March 2012, 00:42:05, Ken KAWAMOTO wrote:
>> Hi,
>>
>> I'm trying to understand how Network.OAuth.Consumer works, and
>> run sample code found here with Twitter API. (the full code is
>> attached at the bottom)
>> http://hackage.haskell.org/packages/archive/hoauth/0.3.3/doc/html/Networ
>> k-OAuth-Consumer.html
>>
>> When I load the test code with GHCi 7.0.2, it raises error:
>>
>> --- Error message begin
>> TestOAuth.hs:21:103:
>>     Ambiguous type variable `m0' in the constraints:
>>       (MonadIO m0) arising from a use of `serviceRequest'
>>                    at TestOAuth.hs:21:103-116
>>       (Monad m0) arising from a use of `>>=' at TestOAuth.hs:21:99-101
>>     Possible cause: the monomorphism restriction applied to the
>> following: response :: m0 Response (bound at TestOAuth.hs:17:1)
>>     Probable fix: give these definition(s) an explicit type signature
>>                   or use -XNoMonomorphismRestriction
>>     In the second argument of `(>>=)', namely
>
>>
>> I seem using NoMonomorphismRestriction is workaround, so I want to add
>> a proper type signature, but I cannot figure out what it is.
>>
>> What confuses me is why GHC didn't just take MonadIO as MonadIO is a
>> Monad according to the definition.
>
> From the definition of response, the two constraints
>
> MonadIO m0
> Monad m0
>
> are inferred. Since response is bound by a simple pattern binding (no
> function arguments) and without a type signature, by the monomorphism
> restriction, it must get a monomorphic type, so the compiler tries to find
> a specific MonadIO (and Monad, but that would follow) to use. But there is
> no defaulting for those constraints (defaulting takes place for constraints
> where at least one of the involved classes is a numeric class [Num,
> Integral, Fractional, ...], and all involved classes are defined in the
> standard libraries), so that fails (no numeric class, and MonadIO is
> defined outside the standard libraries).
>
> To fix it, you just have to define response with a type signature,
>
> response :: MonadIO m => m Response
>
> You should give all your top-level definitions type signatures anyway. It's
> good for documentation, and it gives you much better error messages if you
> make a mistake.



-- 
Ken KAWAMOTO
kentaro.kawamoto at gmail.com