adamconnersax / frames-map-reduce Goto Github PK

I took an exploratory stab at trying to make some of the combinators in this library polymorphic in the interpretation functor and quickly ran up against the limits of its design. I imagine it's not straightforward (having run into similar difficulties myself) to to express folds over records where data may be missing.

Have you given such an implementation any thought?

The expected functionality, I think, would be something like:

• in the case of a key a, use the composite Maybe a for grouping.
• in the case of data as, if there's Nothing, then there's no step, otherwise take the data and step.

The two sources of difficulty in my mind are constraining the payload through the composed functors and, at the same time, running the fold on the undistributed version of the data.

(Separately, it'd also be nice to be able to support ARecs as well.)

cc: @o1lo01ol1o
I finished rounding out the Maybe stuff you did in #2. It's in the latest commit on master. I did the minimum required to restore all the functions that were already there in the non-Maybe version. I added

rgetMaybeField
  :: forall t rs
   . (V.KnownField t, F.ElemOf rs t, V.FieldType (V.Fst t) rs ~ V.Snd t)
  => F.Rec (Maybe :. ElField) rs
  -> Maybe (V.Snd t)
rgetMaybeField = fmap V.getField . V.getCompose . V.rgetf (V.Label @(V.Fst t))

and

-- | Control.Foldl helper for filtering Nothings
maybeFold :: FL.Fold a b -> FL.Fold (Maybe a) b
maybeFold (FL.Fold step begin done) = FL.Fold step' begin done
  where step' x ma = maybe x (step x) ma

to help and to point the way to generalization.

and made your foldAllConstrained into maybeFoldAllConstrained so that I could re-use foldAllConstrained for the case when the input fold is Fold a a which is then generalized to Fold (Maybe a) (Maybe a) via fmap Just . maybeFold

If you have a chance to take a look, that would be great.
Now I'm going to make a branch to try and generalize it all, including the record-type. If that works, I'll try to re-express these as monomorphizations (is that a word??) of the general thing.