Hi Martin,
But the compiler complains about ambiguous type for the writer reult I
am ignoring (message in the above lpaste).
There are a number of types going on here, and the one that's ambiguous is the monoid w in WriterT w m a.
Normally I think the way around this is to provide explicit type
annotation for "foldIt", but in this case the result type depends on the
type of "a" in foldrEntries type and I don't know how to express this
and make the compiler happy.
David's ScopedTypeVariables solution is perfectly cromulent. Here's another one without any pragmas:
foldrEntries ::
(Entry -> a -> a)
-> a
-> (String -> a)
-> Entries
-> a
foldrEntries next done fail' =
isoR . foldrEntriesW (isoL .: next) (isoL done) (isoL . fail')
where
isoL :: a -> WriterT () Identity a
isoL = return
isoR :: WriterT () Identity a -> a
isoR = fst . runIdentity . runWriterT
The (.:) is from Data.Composition.
The isoL and isoR witness the isomorphism, and foldrEntries is written in a point-minimized form that makes it transparent that it's a specialization of foldrEntriesW.
Do eschew verbosities like "return . fst =<< runWriterT". (Hlint should be programmed to catch this. It doesn't yet.) A return that's too near a monadic bind sets off alarms among professional haskell engineers.
(I was able to make it work by calling "tell ()", basically writing a
dummy value, which lets compiler know what the type is, but this is not
so good - I don't want to make artificial function calls like this.)