Skip to content

Commit

Permalink
Give asum, msum, and Product instances explicit kind signatures
Browse files Browse the repository at this point in the history
This adds more explicit `Type -> Type` kind signatures to definitions that
would otherwise be kind-generalized to have overly polymorphic kinds:

* The kinds of `Asum` and `Msum` (the promoted counterparts to the term-level
  `asum` and `msum` functions, respectively).
* The helper type families generated when promoting the `Alternative` and
  `MonadPlus` instances for `Data.Functor.Product`. This sort of kind
  polymorphism isn't observable by users in today's GHC, but it will cause
  problems later in a future version of GHC that implements
  [GHC#23515](https://gitlab.haskell.org/ghc/ghc/-/issues/23515). (See #601.)

(I should have added these as part of #606 or #607, but I forgot them due to an
oversight.)
  • Loading branch information
RyanGlScott committed Jul 1, 2024
1 parent 48b419a commit ae2a94f
Show file tree
Hide file tree
Showing 3 changed files with 69 additions and 10 deletions.
14 changes: 14 additions & 0 deletions singletons-base/CHANGES.md
Original file line number Diff line number Diff line change
Expand Up @@ -55,6 +55,20 @@ next [????.??.??]
`PAlternative` instances for `Compose`. The fact that these instances were so
polymorphic to begin with was an oversight, as these instances could not be
used when `k` was instantiated to any other kind besides `Type`.
* The kinds of `Asum` and `Msum` are less polymorphic than they were before:

```diff
-type Asum :: forall {j} {k} (t :: k -> Type) (f :: j -> k) (a :: j). t (f a) -> f a
+type Asum :: forall (t :: Type -> Type) (f :: Type -> Type) (a :: Type). t (f a) -> f a

-type Msum :: forall {j} {k} (t :: k -> Type) (m :: j -> k) (a :: j). t (m a) -> m a
+type Msum :: forall (t :: Type -> Type) (m :: Type -> Type) (a :: Type). t (m a) -> m a
```

Similarly, the kinds of defunctionalization symbols for these definitions
(e.g., `AsumSym0` and `MSym0`) are less polymorphic. The fact that these were
kind-polymorphic to begin with was an oversight, as these definitions could
not be used when `j` or `k` was instantiated to any other kind besides `Type`.

3.4 [2024.05.12]
----------------
Expand Down
25 changes: 23 additions & 2 deletions singletons-base/src/Data/Foldable/Singletons.hs
Original file line number Diff line number Diff line change
Expand Up @@ -581,16 +581,37 @@ $(singletonsOnly [d|
sequence_ :: (Foldable t, Monad m) => t (m a) -> m ()
sequence_ = foldr (>>) (return ())

-- Note that in the type signatures for `asum` and `msum` below, we explicitly
-- annotate `f` and `m` with the kind (Type -> Type), which is not something
-- that is done in the original base library. This is because when
-- singletons-th promotes type signatures, it omits constraints such as
-- `Alternative f` and `MonadPlus m`, which are essential for inferring that
-- `f` and `m` are of kind `Type -> Type`. Without these constraints, we may
-- end up with a promoted definition that looks like this:
--
-- type Asum :: t (f a) -> f a
--
-- This will result in Asum having a more polymorphic kind than intended,
-- since GHC will generalize Asum's kind to:
--
-- type Asum :: forall {j} {k} (t :: k -> Type) (f :: j -> k) (a :: j). t (f a) -> f a
--
-- Annotating `f :: Type -> Type` (and similarly for `m`) is a clunky but
-- reliable way of preventing this. See also Note [Using standalone kind
-- signatures not present in the base library] in
-- Control.Monad.Singletons.Internal for a similar situation where class
-- definitions can become overly polymorphic unless given an explicit kind.

-- -| The sum of a collection of actions, generalizing 'concat'.
--
-- asum [Just "Hello", Nothing, Just "World"]
-- Just "Hello"
asum :: (Foldable t, Alternative f) => t (f a) -> f a
asum :: forall t (f :: Type -> Type) a. (Foldable t, Alternative f) => t (f a) -> f a
asum = foldr (<|>) empty

-- -| The sum of a collection of actions, generalizing 'concat'.
-- As of base 4.8.0.0, 'msum' is just 'asum', specialized to 'MonadPlus'.
msum :: (Foldable t, MonadPlus m) => t (m a) -> m a
msum :: forall t (m :: Type -> Type) a. (Foldable t, MonadPlus m) => t (m a) -> m a
msum = asum

-- -| The concatenation of all the elements of a container of lists.
Expand Down
40 changes: 32 additions & 8 deletions singletons-base/src/Data/Functor/Product/Singletons.hs
Original file line number Diff line number Diff line change
Expand Up @@ -102,20 +102,44 @@ $(singletonsOnly [d|
Pair f g <*> Pair x y = Pair (f <*> x) (g <*> y)
liftA2 f (Pair a b) (Pair x y) = Pair (liftA2 f a x) (liftA2 f b y)

instance (Alternative f, Alternative g) => Alternative (Product f g) where
empty = Pair empty empty
Pair x1 y1 <|> Pair x2 y2 = Pair (x1 <|> x2) (y1 <|> y2)

instance (Monad f, Monad g) => Monad (Product f g) where
Pair m n >>= f = Pair (m >>= fstP . f) (n >>= sndP . f)
where
fstP (Pair a _) = a
sndP (Pair _ b) = b

instance (MonadPlus f, MonadPlus g) => MonadPlus (Product f g) where
mzero = Pair mzero mzero
Pair x1 y1 `mplus` Pair x2 y2 = Pair (x1 `mplus` x2) (y1 `mplus` y2)

instance (MonadZip f, MonadZip g) => MonadZip (Product f g) where
mzipWith f (Pair x1 y1) (Pair x2 y2) = Pair (mzipWith f x1 x2) (mzipWith f y1 y2)

-- Note that in the instances below, we explicitly annotate `f` with its kind
-- (Type -> Type), which is not something that is done in the original base
-- library. This is because when singletons-th promotes instance declarations,
-- it omits the instance contexts when generating the helper type families.
-- This can lead to the helper type families having overly polymorphic kinds.
-- For example, if the Alternative instance below lacked the explicit
-- (f :: Type -> Type) kind signature, the generated code would look like:
--
-- instance PAlternative (Product f g) where
-- type Empty = EmptyHelper
-- ...
--
-- type EmptyHelper :: forall f g a. Product f g a
--
-- This will result in EmptyHelper having a more polymorphic kind than
-- intended, since GHC will generalize EmptyHelper's kind to:
--
-- type EmptyHelper :: forall {k} (f :: k -> Type) (g :: k -> Type) (a :: k). Product f g a
--
-- Annotating `f :: Type -> Type` is a clunky but reliable way of preventing
-- this. See also Note [Using standalone kind signatures not present in the
-- base library] in Control.Monad.Singletons.Internal for a similar situation
-- where class definitions can become overly polymorphic unless given an
-- explicit kind.
instance (Alternative f, Alternative g) => Alternative (Product (f :: Type -> Type) g) where
empty = Pair empty empty
Pair x1 y1 <|> Pair x2 y2 = Pair (x1 <|> x2) (y1 <|> y2)

instance (MonadPlus f, MonadPlus g) => MonadPlus (Product (f :: Type -> Type) g) where
mzero = Pair mzero mzero
Pair x1 y1 `mplus` Pair x2 y2 = Pair (x1 `mplus` x2) (y1 `mplus` y2)
|])

0 comments on commit ae2a94f

Please sign in to comment.