module Cat.Diagram.Monad where

MonadsπŸ”—

A monad on a category is one way of categorifying the concept of monoid. Specifically, rather than living in a monoidal category, a monad lives in a bicategory. Here, we concern ourselves with the case of monads in the bicategory of categories, so that we may say: A monad is an endofunctor equipped with a unit natural transformation and a multiplication

  record Monad : Type (o βŠ” h) where
    no-eta-equality
    field
      M    : Functor C C
      unit : Id => M
      mult : (M F∘ M) => M

Furthermore, these natural transformations must satisfy identity and associativity laws exactly analogous to those of a monoid.

    field
      left-ident  : βˆ€ {x} β†’ ΞΌ x C.∘ M₁ (Ξ· x) ≑ C.id
      right-ident : βˆ€ {x} β†’ ΞΌ x C.∘ Ξ· (Mβ‚€ x) ≑ C.id
      mult-assoc  : βˆ€ {x} β†’ ΞΌ x C.∘ M₁ (ΞΌ x) ≑ ΞΌ x C.∘ ΞΌ (Mβ‚€ x)

Algebras over a monadπŸ”—

One way of interpreting a monad is as giving a signature for an algebraic theory. For instance, the free monoid monad describes the signature for the theory of monoids, and the free group monad describes the theory of groups.

Under this light, an algebra over a monad is a way of evaluating the abstract operations given by a monadic expression to a concrete value. Formally, an algebra for is given by a choice of object and a morphism

  record Algebra-on (M : Monad) (ob : C.Ob) : Type (o βŠ” h) where
    no-eta-equality
    open Monad M

    field
      Ξ½ : C.Hom (Mβ‚€ ob) ob

This morphism must satisfy equations categorifying those which define a monoid action. If we think of as specifying a signature of effects, then v-unit says that the unit has no effects, and v-mult says that, given two layers it doesn’t matter whether you first join then evaluate, or evaluate twice.

      Ξ½-unit : Ξ½ C.∘ Ξ· ob ≑ C.id
      Ξ½-mult : Ξ½ C.∘ M₁ Ξ½ ≑ Ξ½ C.∘ ΞΌ ob

  Algebra : Monad β†’ Type (o βŠ” h)
  Algebra M = Ξ£ _ (Algebra-on M)

Eilenberg-Moore categoryπŸ”—

If we take a monad as the signature of an (algebraic) theory, and as giving models of that theory, then we can ask (like with everything in category theory): Are there maps between interpretations? The answer (as always!) is yes: An algebra homomorphism is a map of the underlying objects which β€œcommutes with the algebras”.

  record Algebra-hom (M : Monad) (X Y : Algebra M) : Type (o βŠ” h) where
    no-eta-equality
    constructor algebra-hom
    private
      module X = Algebra-on (X .snd)
      module Y = Algebra-on (Y .snd)

    open Monad M

    field
      morphism : C.Hom (X .fst) (Y .fst)
      commutes : morphism C.∘ X.Ξ½ ≑ Y.Ξ½ C.∘ M₁ morphism

  open Algebra-hom

We can be more specific about β€œcommuting with the algebras” by drawing a square: A map in the ambient category is a homomorphism of when the square below commutes.

Since commutes is an identification between morphisms, it inhabits a proposition (because Hom-sets are sets), equality of algebra homomorphisms only depends on an equality of their underlying morphisms.

  Algebra-hom-path
    : {M : Monad} {X Y : Algebra M} {F G : Algebra-hom M X Y}
    β†’ morphism F ≑ morphism G
    β†’ F ≑ G
  Algebra-hom-path x i .morphism = x i
  Algebra-hom-path {M = M} {X} {Y} {F} {G} x i .commutes =
    is-propβ†’pathp (Ξ» i β†’ C.Hom-set _ _ (x i C.∘ X .snd .Algebra-on.Ξ½)
                                      (Y .snd .Algebra-on.Ξ½ C.∘ Monad.M₁ M (x i)))
      (F .commutes) (G .commutes) i

Since the square we drew above commutes for the identity morphism, and we can show that the composite of two algebra homomorphisms is an algebra homomorphism, they assemble into a category: The Eilenberg-Moore category of

  module _ (M : Monad C) where
    private
      module M = Monad M
    open M hiding (M)
    open Precategory
    open Algebra-on

    Eilenberg-Moore : Precategory _ _
    Eilenberg-Moore .Ob = Algebra C M
    Eilenberg-Moore .Hom X Y = Algebra-hom C M X Y

Defining the identity and composition maps is mostly an exercise in categorical yoga:

    Eilenberg-Moore .id {o , x} .morphism = C.id
    Eilenberg-Moore .id {o , x} .commutes =
      C.id C.∘ Ξ½ x     β‰‘βŸ¨ C.id-comm-sym βŸ©β‰‘
      Ξ½ x C.∘ C.id     β‰‘βŸ¨ ap (C._∘_ _) (sym M-id) βŸ©β‰‘
      Ξ½ x C.∘ M₁ C.id  ∎

    Eilenberg-Moore ._∘_ {_ , x} {_ , y} {_ , z} F G .morphism =
      morphism F C.∘ morphism G
    Eilenberg-Moore ._∘_ {_ , x} {_ , y} {_ , z} F G .commutes =
      (morphism F C.∘ morphism G) C.∘ Ξ½ x            β‰‘βŸ¨ C.extendr (commutes G) βŸ©β‰‘
      ⌜ morphism F C.∘ Ξ½ y ⌝ C.∘ M₁ (morphism G)     β‰‘βŸ¨ ap! (commutes F) βŸ©β‰‘
      (Ξ½ z C.∘ M₁ (morphism F)) C.∘ M₁ (morphism G)  β‰‘βŸ¨ C.pullr (sym (M-∘ _ _)) βŸ©β‰‘
      Ξ½ z C.∘ M₁ (morphism F C.∘ morphism G)         ∎
Because we have characterised equality of algebra homomorphisms as equality of their underlying maps, the Eilenberg-Moore category inherits the identity and associativity laws from its underlying category.
    Eilenberg-Moore .idr f = ext (C.idr _)
    Eilenberg-Moore .idl f = ext (C.idl _)
    Eilenberg-Moore .assoc f g h = ext (C.assoc _ _ _)
    Eilenberg-Moore .Hom-set X Y = hlevel 2

By projecting the underlying object of the algebras, and the underlying morphisms of the homomorphisms between them, we can define a functor from Eilenberg-Moore back to the underlying category:

    Forget : Functor Eilenberg-Moore C
    Forget .Fβ‚€ = fst
    Forget .F₁ = Algebra-hom.morphism
    Forget .F-id = refl
    Forget .F-∘ f g = refl

The lemma Algebra-hom-path says exactly that this functor is faithful.

    Forget-is-faithful : is-faithful Forget
    Forget-is-faithful = ext

Free algebrasπŸ”—

In exactly the same way that we may construct a free group by taking the inhabitants of some set as generating the β€œwords” of a group, we can, given an object of the underlying category, build a free on Keeping with our interpretation of monads as logical signatures, this is the syntactic model of with a set of β€œneutrals” chosen from the object

This construction is a lot simpler to do in generality than in any specific case: We can always turn into an by taking the underlying object to be and the algebra map to be the monadic multiplication; The associativity and unit laws of the monad itself become those of the

    Free : Functor C Eilenberg-Moore
    Free .Fβ‚€ A .fst = Mβ‚€ A
    Free .Fβ‚€ A .snd .Ξ½ = mult .Ξ· A
    Free .Fβ‚€ A .snd .Ξ½-mult = mult-assoc
    Free .Fβ‚€ A .snd .Ξ½-unit = right-ident

The construction of free is furthermore functorial on the underlying objects; Since the monadic multiplication is a natural transformation the naturality condition (drawn below) doubles as showing that the functorial action of can be taken as an algebraic action:

    Free .F₁ f .morphism = M₁ f
    Free .F₁ f .commutes = sym $ mult.is-natural _ _ _
    Free .F-id = ext M-id
    Free .F-∘ f g = ext (M-∘ f g)

This is a free construction in the precise sense of the word: it’s the left adjoint to the functor Forget, so in particular it provides a systematic, universal way of mapping from to

    open _⊣_

    Free⊣Forget : Free ⊣ Forget
    Free⊣Forget .unit = NT M.η M.unit.is-natural
    Free⊣Forget .counit .η x =
      record { morphism = x .snd .Ξ½
             ; commutes = sym (x .snd .Ξ½-mult)
             }
    Free⊣Forget .counit .is-natural x y f =
      ext (sym (commutes f))
    Free⊣Forget .zig = ext left-ident
    Free⊣Forget .zag {x} = x .snd .ν-unit