> Is it better to type the derived methods inside the class
> definition or out? Are there any efficiency penalties
> in any of the styles?
>
> In Haskell98 Prelude there is a mixture of both styles, for example,
> (>>) is defined inside the Monad class, but (=<<) is left out.
Good question. The main reason for putting them *in* the class
decl is that it allows the possibility for an instance declaration
to supply a more efficient type-specific implementation of the method.
The down-side is that a client of the class does not know how a method
is implemented, whereas he does know how a function defined outside
the class is implemented.
(>>) is in Monad because conceivably it might be more efficient to
know that the result is not going to be used; but (=<<) is not
becuase it's just an argument-flipped version of (>>=), and so could
not possibly be implemented more efficiently. But arguably sequence and
friends should be in Monad too? It's hard to tell.
Bottom line: it's a judgement call
Simon
