Implementing Syntax

Every algebra should have associated syntax. Implementing syntax is a little bit more involved than defining an algebra due to the extra work we do to make type inference work nicely. The goal is to build up the algebra required by a Picture based on the operations used. For example, notice the inferred type of the Picture below.

import doodle.core.*
import doodle.syntax.all.*
circle(100).beside(square(20)).strokeColor(Color.red)

This type reflects exactly the algebras used in constructing the Picture. For users working with a single backend this is never an issue as they work from the constructors on the Picture object which start with all the algebras supported by the backend. However for users working across backends this is essential to avoid a lot of type juggling.

There are different patterns for syntax for constructors and combinators. Syntax should always be defined inside a trait that is mixed into the relevant all object.

Constructor Syntax

Constructor syntax is simply a method that produces a Picture with the relevant type. For example, if we have a constructor called square which relates to an algebra called Shape, we can implement syntax as

def square(width: Double): Picture[Shape, Unit] =
  new Picture[Shape, Unit] {
    def apply(implicit algebra: Shape): algebra.Drawing[Unit] =
      algebra.square(width, height)
  }

Combinator Syntax

The pattern for implementing combinator syntax is more involved, as we must worry about type inference. Here's the pattern:

  1. Start with an extension extending a Picture with a polymorphic Algebra type parameter

    extension ExampleOps[Alg <: Algebra, A](picture: Picture[Alg, A]) {
  2. Methods on the extension return a Picture with additional algebras that the method requires

     def strokeColor(color: Color): Picture[Alg with Style] = ???
    }

Binary operations, such as on, require two polymorphic Algebra type parameters. Here's the implementation of on showing this (Alg and Alg2).

implicit class LayoutPictureOps[Alg <: Algebra, A](
    picture: Picture[Alg, A]
) {
  def on[Alg2 <: Algebra](
      that: Picture[Alg2, A]
  )(implicit s: Semigroup[A]): Picture[Alg with Alg2 with Layout, A] =
    new Picture[Alg with Alg2 with Layout, A] {
      def apply(implicit
          algebra: Alg with Alg2 with Layout
      ): algebra.Drawing[A] =
        algebra.on(picture(algebra), that(algebra))
    }
}

Implementing a Backend→