Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add failing for traversals #393

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Add failing for traversals #393

wants to merge 1 commit into from

Conversation

arybczak
Copy link
Collaborator

@arybczak arybczak commented Jan 18, 2021

We have adjoin, might as well add failing.

The awkwardness here is about naming, since this PR renames existing Fold equivalents.

There is also an issue that afailing for traversals is safe, while failing for traversals is safe only on disjoint traversals.

Perhaps we want a different name than failing to avoid a minor API breakage (similar to how we have summing for folds and adjoin for traversals). I have no idea what is a good name though.

@arybczak
Copy link
Collaborator Author

arybczak commented Feb 8, 2021

@phadej @adamgundry thoughts?

@georgefst
Copy link
Contributor

georgefst commented Sep 29, 2021

I just spent quite a while looking for this, while porting some lens code that just used (<>) with AffineTraversals (EDIT: with Folds inlens's case, but while porting they become AffineTraversals, and need to be affine in order to use ^?) (in the end, the existing afailing turned out to be just about good enough for my use case, for now).

I'd suggest that this PR should also add a Monoid structure section to the docs for AffineTraversal, like Fold, AffineFold and Traversal have. PS. there are minor inconsistencies in those three existing sections in the title and URL.

@cdfa
Copy link

cdfa commented Feb 15, 2022

I would like to see failing variations for other optics too! However, instead of coming up with clever name variations, maybe we can generalize it into one gfailing (and igfailing for indexed optics).
Additionally, it would be nice if the kind of the "recovery" optic would be preserved in some cases. For example, combining in affine fold with a getter could yield a getter.
I had a go at this and you can find the result below.

Additionally, I think we could refine the precondition that the combined optics are disjoint a little more to allow for things like ix i `gfailing` _last.
The intuition is that k `gfailing` l should be legal when the success/failure of k is independent of the values of l's foci.
More accurately stated: k `gfailing` l is legal when k and l are disjoint or l focuses at least as "deep" as k.

Optics.Fallible
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FunctionalDependencies #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE TypeApplications #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE UndecidableInstances #-}

module Optics.Fallible where

import Data.Type.Equality

import Optics
import Optics.Internal.Utils

class FallibleOptic k l m s t a b | k l -> m where
    {-|
    NB: @gfailing k l@ only produces legal R/W optics when the foci of @k@ and @l@ are disjoint or @l@ focuses at least as "deep" as @k@.
    For example, @_tail \`gfailing\` simple@ is illegal, because composition is not preserved, i.e. with @o = _tail \`gfailing\` simple@, @Identity . cons 1@ and @g = Identity . cons 2@, we have
    
    @
    getCompose $ traverseOf o (Compose . fmap f . g) [] ≡ Identity (Identity [ 1 , 2 ])
        /= Identity (Identity [ 2 , 1 ]) ≡ fmap (traverseOf o f) $ traverseOf o g []
    @

    On the other hand, @ix i \`gfailing\` _last@ would be legal.
    In other words, the success/failure of @k@ cannot depend on the values of @l@'s foci.
    -}
    gfailing
        :: Optic k is s t a b -> Optic l is s t a b -> Optic m NoIx s t a b

infixl 3 `gfailing` -- Same as (<|>)

-- Prism
instance FallibleOptic A_Prism An_Iso A_Lens s t a b where
    gfailing k l = gfailing k $ castOptic @A_Lens l

instance FallibleOptic A_Prism A_Lens A_Lens s t a b where
    gfailing k l = gfailing (castOptic @An_AffineTraversal k) l

instance FallibleOptic A_Prism A_Prism An_AffineTraversal s t a b where
    gfailing = gfailing `on` castOptic @An_AffineTraversal

instance FallibleOptic A_Prism A_Getter A_Getter s s a a where
    gfailing k l = gfailing (castOptic @An_AffineFold k) l

-- AffineTraversal
instance FallibleOptic An_AffineTraversal A_Lens A_Lens s t a b where
    gfailing k l = withAffineTraversal k $ \matchK _ -> withLens l $ \viewL _ ->
        lens (\s -> fromRight (viewL s) $ matchK s)
             (\s b -> fromMaybe (set l b s) $ failover k (const b) s)

instance FallibleOptic An_AffineTraversal An_AffineTraversal An_AffineTraversal s t a b where
    gfailing k l = atraversalVL $ \point f s -> let
        OrT visited fu = atraverseOf k (OrT False . point) (wrapOrT . f) s
        in if visited then fu else atraverseOf l point f s

-- AffineFold
instance FallibleOptic An_AffineFold A_Getter A_Getter s s a a where
    gfailing k l = to $ \s -> fromMaybe (view l s) $ preview k s

instance FallibleOptic An_AffineFold An_AffineFold An_AffineFold s s a a where
    gfailing k l = afolding $ \s -> preview l s <|> preview k s

-- Traversal
instance FallibleOptic A_Traversal A_Traversal A_Traversal s t a b where
    gfailing k l = traversalVL $ \f s -> let
        OrT visited fu = traverseOf k (wrapOrT . f) s
        in if visited then fu else traverseOf l f s

-- Fold
instance FallibleOptic A_Fold A_Fold A_Fold s s a a where
    gfailing k l = foldVL $ \f s -> let
        OrT visited fu = traverseOf_ k (wrapOrT . f) s
        in if visited then fu else traverseOf_ l f s

instance {-# OVERLAPPABLE #-}( JoinKinds k l m
                             , Is k m
                             , Is l m
                             , FallibleOptic m m m s t a b
                             , (k == l) ~ 'False
                             ) => FallibleOptic k l m s t a b where
    gfailing k l = gfailing (castOptic @m k) (castOptic @m l)

I can make a full PR if you're interested.

@Lev135
Copy link

Lev135 commented Jun 9, 2023

There is also an issue that afailing for traversals is safe, while failing for traversals is safe only on disjoint traversals.

Maybe it should be named unsafeAfailing like unsafeFiltered?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

4 participants