Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Variance & Types - Scalabase

Variance & Types - Scalabase

What is Variance, the connection between Types & Variance and how Scala supports it?
Deep dive into types of Variances and how they relate to Functions.

Asjad Baig

May 14, 2021

Transcript

64 slides
  1. Will discuss... • Type Bounds • How to use ::

    Variance and its type • Liskov Substitution Principle • Function trait connection with Variance • How compiler ensures Variance
  2. Type Bounds 1. A & B both are types 2.

    A .: B A subtype of B 3. A .: B B subtype of A / A supertype of B
  3. Lower Bounds • [A .: Bulldog] • Range of possibilities

    A can have ? • {Dog, Animal, Any, Bulldog}
  4. Upper Bounds • [A .: Dog] • Range of possibilities

    A can have ? • {Dobermann, Greyhound, Bulldog, Nothing, Dog}
  5. Multi Bounds • [Animal .: A .: Bulldog] • Range

    of possibilities A can have ? • {Animal, Dog, Bulldog}
  6. Variance • Scala supports subtyping • Variance refers to how

    subtyping between more complex types relates to subtyping between their components
  7. CoVariance • C[A] .: C[B] • The subtyping relation is

    preserved in the same direction • Parameterize as C[+ T] (Type Notation) • Most Immutable Scala Collections are Covariant
  8. ContraVariance • C[A] .> C[B] • The subtyping relation is

    in the opposite direction • Parameterize as C[- T]
  9. Modelling ZooKeeper • class ZooKeeper[-T] • Is there a way

    to restrict it to just Animals ? • class ZooKeeper[-T .: Animal]
  10. InVariance • C[A] .= C[B] • There’s no relation with

    the parameterized type • Parameterize as C[ T] • Mutable collections are Invariant
  11. How it started • Let functionQ(x) be a property provable

    about objects x of type B . • Then functionQ(y) should be provable for objects y of type A where A .: B
  12. How it started If A .: B, then everything one

    can to do with a value of type B one should also be able to do with a value of type A. How it’s going • Let functionQ(x) be a property provable about objects x of type B . • Then functionQ(y) should be provable for objects y of type A where A .: B
  13. If A .: B, then everything one can to do

    with a value of type B one should also be able to do with a value of type A. How it’s going
  14. How it started • Let functionQ(x) be a property provable

    about objects x of type B . • Then functionQ(y) should be provable for objects y of type A where A .: B
  15. Function Typing Rule • If we have functions ◦ fn1

    = A1 .> B1 ◦ fn2 = A2 .> B2 • A1 .> B1 .: A2 .> B2 (i.e fn1 .: fn2) • Only when :: ◦ B1 .: B2 Covariant in Result Types ◦ A1 .: A2 Contravariant in Argument types
  16. Function for Elephant • Elephant eats Grass • fn1 ::

    Animal .> Napier • fn2 :: Elephant .> Grass
  17. Function Trait • How N input function should look ?

    • FunctionN[ -Input1,..,-InputN, +Result] • But there’s a limitation here with value N
  18. Function Trait • How N input function should look ?

    • FunctionN[ -Input1,..,-InputN, +Result] • But there’s a limitation here with value N • N = 22
  19. Function Trait • How N input function should look ?

    • FunctionN[ -Input1,..,-InputN, +Result] • But there’s a limitation here with value N • N = 22 • But, in scala3 this limitation is no more!!
  20. Scala Compiler Checks • Covariant type params can only appear

    in method results • Contravariant type param can only appear in method param • Invariant type param can appear anywhere
  21. Inference 1. Scala immutable collection, Producer => Covariant 2. Consumer

    modelling => Contravariant 3. Mutable Collections => Invariant 4. Thanks to Bounds => i.e. add() operation for Covariant types List 5. FunctionN [ -Input1,..,-InputN , +Result] (scala3)
  22. Resources 1. FP Principles EPFL : Coursera 2. Foundations of

    FP : Julien Truffaut 3. RockTheJVM 4. DevInsideYou

Other Decks in Programming