22 Jun
2010
22 Jun
'10
6:32 a.m.
On Jun 19, 2010, at 1:48 PM, Heinrich Apfelmus wrote:
In my code, mzero is indeed an identity for orElse [...] The observation is any action can be brought into one of the forms
mzero return a `mplus` return b `mplus` ...
which corresponds to the list type [a] .
Ok that makes sense because the list types supports both a cancelling `orElse` and a distributive `mplus` with identity `mzero`.
[...] you can implement your type as
newtype CMaybe a = CMaybe { forall b . (a -> [b]) -> [b] }
Yes. For me it was interesting to see how far we get by wrapping `Maybe` in `Codensity`: we get more than `Maybe` but not as much as `[]`. Sebastian -- Underestimating the novelty of the future is a time-honored tradition. (D.G.)