open import 1Lab.Path.IdentitySystem
open import 1Lab.Path.Reasoning
open import 1Lab.Path.Groupoid
open import 1Lab.Type.Sigma
open import 1Lab.Underlying hiding (Σ-syntax)
open import 1Lab.Univalence
open import 1Lab.Type.Pi
open import 1Lab.HLevel
open import 1Lab.Equiv
open import 1Lab.Path
open import 1Lab.Type

module 1Lab.Type.Pointed where

Pointed types🔗

A pointed type is a type $A$ equipped with a choice of base point $∙_{A} .$

Type∙ : ∀ ℓ → Type (lsuc ℓ)
Type∙ ℓ = Σ (Type ℓ) (λ A → A)

private variable
  ℓ ℓ' : Level
  A B C : Type∙ ℓ

If we have pointed types $(A, a)$ and $(B, b),$ the most natural notion of function between them is not simply the type of functions $A \to B,$ but rather those functions $A \to B$ which preserve the basepoint, i.e. the functions $f : A \to B$ equipped with paths $f (a) \equiv b .$ Those are called pointed maps.

_→∙_ : Type∙ ℓ → Type∙ ℓ' → Type _
(A , a) →∙ (B , b) = Σ[ f ∈ (A → B) ] (f a ≡ b)

The type of pointed maps $A \to_{*} (B, b)$ is always inhabited by the constant function with value $b,$ which we refer to as the zero map. This makes the type $A \to_{*} B$ itself a pointed type; in the code, we write Maps∙. As further examples, the identity is evidently pointed, and the composition of pointed maps is once again pointed.

zero∙ : A →∙ B
zero∙ {B = _ , b₀} = record { snd = refl }

id∙ : A →∙ A
id∙ = id , refl

_∘∙_ : (B →∙ C) → (A →∙ B) → A →∙ C
(f , f*) ∘∙ (g , g*) = record
  { fst = f ∘ g
  ; snd = ap f g* ∙ f*
  }

Maps∙ : Type∙ ℓ → Type∙ ℓ' → Type∙ (ℓ ⊔ ℓ')
Maps∙ A B .fst = A →∙ B
Maps∙ A B .snd = zero∙

infixr 0 _→∙_
infixr 40 _∘∙_

The product of pointed types is naturally pointed (at the pairing of the basepoints), and this definition makes the fst and snd projections into pointed maps.

_×∙_ : Type∙ ℓ → Type∙ ℓ' → Type∙ (ℓ ⊔ ℓ')
(A , a) ×∙ (B , b) = A × B , a , b

infixr 5 _×∙_

fst∙ : A ×∙ B →∙ A
fst∙ = fst , refl

snd∙ : A ×∙ B →∙ B
snd∙ = snd , refl

Paths between pointed maps are characterised as pointed homotopies. If we are comparing $(f, φ)$ and $(g, γ)$ as pointed maps $(A, a_{0}) \to_{*} (B, b_{0}),$ it suffices to give a homotopy $h : f \equiv g$ and a Square with the boundary below. We note in passing that this is equivalent to asking for a proof that $φ \equiv h (a_{0}) \cdot γ .$

`agda module _ {ℓ ℓ'} {A@(_ , a₀) : Type∙ ℓ} {B@(_ , b₀) : Type∙ ℓ'} {f∙@(f , φ) g∙@(g , γ) : A →∙ B} where

  funext∙ : (h : ∀ x → f x ≡ g x) → Square (h a₀) φ γ refl → f∙ ≡ g∙
  funext∙ h pth i = record
    { fst = funext h i
    ; snd = pth i
    }

  _ : (h : ∀ x → f x ≡ g x) → φ ≡ h a₀ ∙ γ → f∙ ≡ g∙
  _ = λ h α → funext∙ h (flip₁ (∙→square' α))

Pointed equivalences🔗

Combining our univalent understanding of paths in the universe, and our understanding of paths in $Σ$ types, it stands to reason that identifications $(A, a_{0}) \equiv (B, b_{0})$ in the space of pointed types are given by equivalences $e : A ≃ B$ which carry $a_{0}$ to $b_{0} .$ We call these a pointed equivalence from $(A, a_{0})$ to $(B, b_{0});$ and, as expected, we can directly use cubical primitives to turn a pointed equivalence into a path of pointed types.

_≃∙_ : Type∙ ℓ → Type∙ ℓ' → Type _
(A , a₀) ≃∙ (B , b₀) = Σ[ e ∈ A ≃ B ] (e · a₀ ≡ b₀)

ua∙ : A ≃∙ B → A ≡ B
ua∙ (e , p) i .fst = ua e i
ua∙ (e , p) i .snd = path→ua-pathp e p i

Of course, a pointed equivalence has an underlying pointed map, by simply swapping the quantifiers. Less directly, the inverse of a pointed equivalence is a pointed equivalence as well.

module Equiv∙ {ℓ ℓ'} {A@(_ , a₀) : Type∙ ℓ} {B@(_ , b₀) : Type∙ ℓ'} (e : A ≃∙ B) where

  to∙ : A →∙ B
  to∙ = e .fst .fst , e .snd

  open Equiv (e .fst) hiding (inverse) public

  inverse : B ≃∙ A
  inverse .fst = Equiv.inverse (e .fst)
  inverse .snd = injective $
    e · from b₀ ≡⟨ Equiv.ε (e .fst) _ ⟩
    b₀          ≡˘⟨ e .snd ⟩
    e · a₀      ∎

  from∙ : B →∙ A
  from∙ = inverse .fst .fst , inverse .snd

id≃∙ : ∀ {ℓ} {A : Type∙ ℓ} → A ≃∙ A
id≃∙ = id≃ , refl

≃∙⟨⟩-syntax : ∀ {ℓ ℓ₁ ℓ₂} (A : Type∙ ℓ) {B : Type∙ ℓ₁} {C : Type∙ ℓ₂}
            → B ≃∙ C → A ≃∙ B → A ≃∙ C
≃∙⟨⟩-syntax A (g , pt) (f , pt') = f ∙e g , ap (g .fst) pt' ∙ pt

_≃∙˘⟨_⟩_ : ∀ {ℓ ℓ₁ ℓ₂} (A : Type∙ ℓ) {B : Type∙ ℓ₁} {C : Type∙ ℓ₂}
        → B ≃∙ A → B ≃∙ C → A ≃∙ C
A ≃∙˘⟨ f ⟩ g = ≃∙⟨⟩-syntax _ g (Equiv∙.inverse f)

_≃∙⟨⟩_ : ∀ {ℓ ℓ₁} (A : Type∙ ℓ) {B : Type∙ ℓ₁} → A ≃∙ B → A ≃∙ B
x ≃∙⟨⟩ x≡y = x≡y

_≃∙∎ : ∀ {ℓ} (A : Type∙ ℓ) → A ≃∙ A
x ≃∙∎ = id≃∙

infixr 2 ≃∙⟨⟩-syntax _≃∙⟨⟩_ _≃∙˘⟨_⟩_
infix  3 _≃∙∎
infix 21 _≃∙_

syntax ≃∙⟨⟩-syntax x q p = x ≃∙⟨ p ⟩ q

path→equiv∙ : A ≡ B → A ≃∙ B
path→equiv∙ p .fst = path→equiv (ap fst p)
path→equiv∙ p .snd = from-pathp (ap snd p)

ua∙-id-equiv : ua∙ {A = A} id≃∙ ≡ refl
ua∙-id-equiv {A = A , a₀} i j .fst = ua-id-equiv {A = A} i j
ua∙-id-equiv {A = A , a₀} i j .snd = attach (∂ j ∨ i)
  (λ { (j = i0) → a₀ ; (j = i1) → a₀ ; (i = i1) → a₀ })
  (inS a₀)

Moreover, we can prove that pointed equivalences are an identity system on the type of pointed types, again by a pretty direct cubical argument.

univalence∙-identity-system
  : is-identity-system {A = Type∙ ℓ} _≃∙_ (λ _ → id≃∙)
univalence∙-identity-system .to-path = ua∙
univalence∙-identity-system .to-path-over {a = A , a₀} (e , e*) i = record
  { fst =
    let
      f : ∀ i → A → ua e i
      f i a = path→ua-pathp e refl i
    in f i , is-prop→pathp (λ i → is-equiv-is-prop (f i)) id-equiv (e .snd) i
  ; snd = λ j → path→ua-pathp e (λ k → e* (j ∧ k)) i
  }

Note that this immediately lets us obtain a pointed equivalence induction principle.

Equiv∙J
  : ∀ {ℓ ℓ'} {A : Type∙ ℓ} (P : (B : Type∙ ℓ) → A ≃∙ B → Type ℓ')
  → P A id≃∙
  → ∀ {B} e → P B e
Equiv∙J = IdsJ univalence∙-identity-system

Homogeneity🔗

Coming up with pointed homotopies is tricky when there’s a lot of path algebra involved in showing that the underlying functions are identical, since we would have to trace the pointedness witness along this construction. However, there is a simplifying assumption we can impose on the codomain that makes this much simpler, allowing us to consider only the underlying functions to begin with.

We say that a type $A$ is homogeneous if, given two choices of basepoint $a_{0}, a_{1} : A,$ we have that $(A, a_{0}) = (A, a_{1})$ in the type of pointed types.

Homogeneous : Type ℓ → Type _
Homogeneous A = ∀ {x y} → Path (Type∙ _) (A , x) (A , y)

By univalence, this is equivalent to asking for, given $a_{0}, a_{1},$ an equivalence $e : A ≃ A$ with $e (a_{0}) = a_{1} .$ This provides us some example that the notion is not vacuous: for example, if we can decide equality in $A,$ then we can build such an equivalence by case analysis.

Another example is readily given by path spaces, where if we are given $p, q : x \equiv y$ then precomposition with $q \cdot p^{- 1}$ is an auto-equivalence of $x \equiv y$ which sends $p$ to $q,$ thence an identification between the pointed types $(x \equiv y, p)$ and $(x \equiv y, q) .$

instance
  Path-homogeneous : ∀ {ℓ} {A : Type ℓ} {x y : A} → Homogeneous (Path A x y)
  Path-homogeneous {x = _} {_} {p} {q} = ua∙ record
    { fst = ∙-pre-equiv (q ∙ sym p)
    ; snd = ∙-cancelr q p
    }

If $f, g : A \to_{*} B$ are pointed maps into a homogeneous type $B,$ then to get an identity $f \equiv g$ it suffices to identify the underlying unpointed maps $f, g : A \to B .$

homogeneous-funext∙
  : ∀ {ℓ ℓ'} {A : Type∙ ℓ} {B : Type∙ ℓ'}
  → ⦃ _ : Homogeneous (B .fst) ⦄
  → ∀ {f g : A →∙ B} (h : ∀ x → f .fst x ≡ g .fst x)
  → f ≡ g
homogeneous-funext∙ {A = A} {B = B , b₀} {f = f∙@(f , f*)} {g∙@(g , g*)} h =
  subst (λ b → PathP (λ i → A →∙ b i) f∙ g∙) fix bad where
    hom : ∀ x → Path (Type∙ _) (B , b₀) (B , x)
    hom x = auto

    fg* = sym f* ∙∙ h (A .snd) ∙∙ g*

    bad : PathP (λ i → A →∙ B , fg* i) f∙ g∙
    bad i .fst x = h x i
    bad i .snd j = ∙∙-filler (sym f*) (h (A .snd)) g* j i

    fix : Square {A = Type∙ _} refl (λ i → B , fg* i) refl refl
    fix i j = hcomp (∂ i ∨ ∂ j) λ where
      k (k = i0) → B , b₀
      k (i = i0) → hom (fg* j) k
      k (i = i1) → hom b₀ k
      k (j = i0) → hom b₀ k
      k (j = i1) → hom b₀ k