Polymorphically deriving PLift,PConstant without the headaches

[t1lde]

DerivePConstantViaNewtype is pretty painful to use with polymorphic types - the Constraints required for the derived instances are leaky. An example is:

instance 
  forall a. 
  ( a ~ (PConstanted (PLifted a))
  , PUnsafeLiftDecl a
  ) =>
  PUnsafeLiftDecl (PFoo a) where 
  type PLifted (PFoo a) = Foo tag (PLifted a)

deriving via
  (DerivePConstantViaNewtype 
    (Foo  a)
    (PFoo (PConstanted a))
    (PConstanted a))
  instance 
  (a ~ (PLifted (PConstanted a))
  , PUnsafeLiftDecl (PConstanted a)
  , PConstant a
  ) =>
  PConstant (Foo a) 

One idea to simplify is with Constraint synonyms which provide a shorthand for the constraints:

type PConstantable :: Constraint
type PConstantable a =  (a ~ (PLifted (PConstant a)), PUnsafeLiftDecl (PConstanted a), PConstant a)

type PLiftable a :: Constraint
type PLiftable a = (a ~ (PConstanted (PLifted a)), PUnsafeLiftDecl a)

[peter-mlabs]

As an example of using Data instead, we can have

data Component a = Component
  {cValue :: Plutus.Value,
   cDatum :: a 
  }

data PComponent (a :: PType)  (s :: S)
  = PComponent (Term s (PDataRecord '["_0" ':= PValue, "_1" ':= a]))
  deriving stock (GHC.Generic)
  deriving anyclass (SOP.Generic, PIsDataRepr)
  deriving
    (PMatch, PIsData)
    via PIsDataReprInstances (PComponent a)

instance 
  forall a . 
  ( a ~ (PConstanted (PLifted a))
  , PUnsafeLiftDecl a
  ,  Ledger.FromData (Component (PLifted a))
  , Ledger.ToData (Component (PLifted a))
  ) =>
  PUnsafeLiftDecl (PComponent a) where 
  type PLifted (PComponent a) = Component (PLifted a) 


deriving via
  (DerivePConstantViaData  
    (Component a)
    (PComponent (PConstanted a)))
  instance 
  (a ~ (PLifted (PConstanted a))
  , PUnsafeLiftDecl (PConstanted a)
  , PConstant a
  , Ledger.FromData (Component a)
  , Ledger.ToData (Component a)
  ) =>
  PConstant (Component a)

[L-as]

For type PLiftable a = (a ~ (PConstanted (PLifted a)), PUnsafeLiftDecl a), you should just make PLift a do this @t1lde . This ought to be enough in 99% of cases.

[L-as]

@peter-mlabs if you want to you can try contributing the PR