julianpeeters / polynomial   0.6.0

Apache License 2.0 GitHub

The category of Poly, simply typed.

Scala versions: 3.x
Scala.js versions: 1.x
Scala Native versions: 0.4

polynomial

Based on the polynomial functors described in Niu and Spivak


Add the dependencies:

  • libraries for Scala 3.4+ (JS, JVM, and Native platforms)
  • mermaid integration (optional)
"com.julianpeeters" %% "polynomial" % "0.6.0"         // required
"com.julianpeeters" %% "polynomial-mermaid" % "0.6.0" // optional

polynomial

The polynomial library provides the following implementation of poly:

  • objects: built-in ADTs for monomial, binomial, and trinomial Store and Interface functors
  • morphisms: PolyMap, or ~>, a natural transformation between polynomial functors
  • products:
    • Cartesian, or ×, a categorical product implemented as match types
    • Composition, or , a substitution product implemented as match types
    • Tensor, or , a parallel product implemented as match types
import polynomial.`object`.*
import polynomial.morphism.~>

// Examples
type `2y⁵¹²`           = Monomial.Interface[(Byte, Boolean), Boolean, _]
type `y² + 2y`         = Binomial.BiInterface[Boolean, Unit, Unit, Boolean, _]
type `2y²`             = Monomial.Store[Boolean, _]
type `0`               = Monomial.Interface[Nothing, Nothing, _]
type `1`               = Monomial.Interface[Unit, Nothing, _]
type `y² + 2y → 2y⁵¹²` = (`y² + 2y` ~> `2y⁵¹²`)[_]

FAQ

Q: What are we losing by using simple types rather than dependent types?

A: Simple types can easily model monomial lenses, but they are not flexible enough to model fully dependent lenses.

However, a rich subset of dependent lenses can be implemented, under the following constraints:

  • the positions and directions of the polynomial are related by an ADT
  • the number of terms in the polynomial is equal to the number of members of the ADT

For example, Bi lens can be pameterized by Option such that its terms are exponentiated by Some[A] and None.type, and behaves as a dual-channeled monomial lens.

polynomial-mermaid

Certain lenses can be interpreted graphically. Given a Mermaid instance for a PolyMap, a mermaid flowchart definition can be printed, with titles and labels in the following formats:

  • Cardinal: render exponents and coefficients as integer values
  • Custom: render custom labels for variables, exponents and coefficients
  • Generic: render exponents and coefficients as, e.g., A instead of a Byte
  • Specific: render exponents and coefficients as, e.g., Byte instead of a A
import polynomial.`object`.Monomial.{Interface}
import polynomial.mermaid.{Format, Mermaid, given}
import polynomial.morphism.~>

type F[Y] = (Interface[Byte, Char, _] ~> Interface[Byte, Char, _])[Y]

val M: Mermaid[F] = summon[Mermaid[F]]
// M: Mermaid[F] = polynomial.mermaid.Mermaid$$anon$3@49cbc7e8

println(M.showGraph(graphFmt = Format.Generic))
// ```mermaid
// graph LR;
//   A1:::hidden---|<span style="font-family:Courier">A<sub>2</sub></span>|A_B1[ ]:::point
// subgraph s[ ]
//   A_B1:::point---QB1
//   QB1[<span style="font-family:Courier">B<sub>1</sub></span>𝑦<sup><span style="font-family:Courier">A<sub>1</sub></span></sup>]:::empty
//   QB1---B_B1
// end
// B_B1[ ]:::point---|<span style="font-family:Courier">B<sub>2</sub></span>|B1:::hidden;
// 
// classDef empty fill:background;
// classDef point width:0px, height:0px;
// classDef title stroke-width:0px, fill:background;
// ```
graph LR;
  A:::hidden---|Byte|S[Boolean]---|Char|B:::hidden;

classDef empty fill:background;
classDef point width:0px, height:0px;
classDef title stroke-width:0px, fill:background;

(Note: GitHub currently ignores formatting, please use mermaid.live)