[Haskell-cafe] type-class inference

Brandon Allbery allbery.b at gmail.com
Sat Aug 13 01:29:37 CEST 2011

On Fri, Aug 12, 2011 at 19:08, Brandon Allbery <allbery.b at gmail.com> wrote:

> On Fri, Aug 12, 2011 at 18:52, Patrick Browne <patrick.browne at dit.ie>wrote:
>> Why does the Haskell :type command only sometimes print the type-class?
> Haskell infers the most specific type applicable.  If the most specific it
> can get is a typeclass, that's what it produces; if it can infer an explicit
> type, it will.

By the way, a possible source of confusion here is the combination of the
monomorphism restriction and defaulting, especially GHCi's extended
defaulting.  The monomorphism restriction says that if you don't provide a
way to *easily* infer a type for a binding (in practice this means there are
no parameters), Haskell insists that the binding is not polymorphic unless
you explicitly provide a type signature. Defaulting is how it accomplishes
this:  there is a list of default types that can be applied when a concrete
type is required and none is available, and the first one that typechecks is
used.  The Haskell standard specifies Double and Integer as default types;
GHCI's extended defaulting (or GHC in general with -XExtendedDefaultRules)
adds () aka "unit".  So, for example, something that you might expect to be
(Num a => a) may end up being Integer due to the monomorphism restriction
requiring a concrete type and defaulting supplying one.

(There's a widely expressed sentiment that the monomorphism restriction
should go away because the confusion it engenders is worse than the problems
it solves; on the other hand, GHC recently added a new application of it
(monomorphic pattern bindings).  In any case, if you want to play around
with types in GHCi, you may want to ":set -XNoMonomorphismRestriction
-XNoMonoPatBinds" so you can see how types actually behave in the wild.)

brandon s allbery                                      allbery.b at gmail.com
wandering unix systems administrator (available)     (412) 475-9364 vm/sms
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.haskell.org/pipermail/haskell-cafe/attachments/20110812/2e79f62a/attachment.htm>

More information about the Haskell-Cafe mailing list