------------------------------------------------------------------------
-- The Agda standard library
--
-- Some Vec-related properties
------------------------------------------------------------------------

module Data.Vec.Properties where

open import Algebra
open import Category.Applicative.Indexed
open import Category.Monad
open import Category.Monad.Identity
open import Data.Vec
open import Data.Nat
open import Data.Empty using (⊥-elim)
import Data.Nat.Properties as Nat
open import Data.Fin as Fin using (Fin; zero; suc; toℕ; fromℕ)
open import Data.Fin.Props using (_+′_)
open import Function
open import Function.Inverse using (_↔_)
open import Relation.Binary

module UsingVectorEquality {s₁ s₂} (S : Setoid s₁ s₂) where

  private module SS = Setoid S
  open SS using () renaming (Carrier to A)
  import Data.Vec.Equality as VecEq
  open VecEq.Equality S

  replicate-lemma : ∀ {m} n x (xs : Vec A m) →
                    replicate {n = n}     x ++ (x ∷ xs) ≈
                    replicate {n =  + n} x ++      xs
  replicate-lemma zero    x xs = refl (x ∷ xs)
  replicate-lemma (suc n) x xs = SS.refl ∷-cong replicate-lemma n x xs

  xs++[]=xs : ∀ {n} (xs : Vec A n) → xs ++ [] ≈ xs
  xs++[]=xs []       = []-cong
  xs++[]=xs (x ∷ xs) = SS.refl ∷-cong xs++[]=xs xs

  map-++-commute : ∀ {b m n} {B : Set b}
                   (f : B → A) (xs : Vec B m) {ys : Vec B n} →
                   map f (xs ++ ys) ≈ map f xs ++ map f ys
  map-++-commute f []       = refl _
  map-++-commute f (x ∷ xs) = SS.refl ∷-cong map-++-commute f xs

open import Relation.Binary.PropositionalEquality as P
  using (_≡_; _≢_; refl; _≗_)
open import Relation.Binary.HeterogeneousEquality using (_≅_; refl)

-- lookup is an applicative functor morphism.

lookup-morphism :
  ∀ {a n} (i : Fin n) →
  Morphism (applicative {a = a})
           (RawMonad.rawIApplicative IdentityMonad)
lookup-morphism i = record
  { op      = lookup i
  ; op-pure = lookup-replicate i
  ; op-⊛    = lookup-⊛ i
  }
  where
  lookup-replicate : ∀ {a n} {A : Set a} (i : Fin n) →
                     lookup i ∘ replicate {A = A} ≗ id {A = A}
  lookup-replicate zero    = λ _ → refl
  lookup-replicate (suc i) = lookup-replicate i

  lookup-⊛ : ∀ {a b n} {A : Set a} {B : Set b}
             i (fs : Vec (A → B) n) (xs : Vec A n) →
             lookup i (fs ⊛ xs) ≡ (lookup i fs $ lookup i xs)
  lookup-⊛ zero    (f ∷ fs) (x ∷ xs) = refl
  lookup-⊛ (suc i) (f ∷ fs) (x ∷ xs) = lookup-⊛ i fs xs

-- tabulate is an inverse of flip lookup.

lookup∘tabulate : ∀ {a n} {A : Set a} (f : Fin n → A) (i : Fin n) →
                  lookup i (tabulate f) ≡ f i
lookup∘tabulate f zero    = refl
lookup∘tabulate f (suc i) = lookup∘tabulate (f ∘ suc) i

tabulate∘lookup : ∀ {a n} {A : Set a} (xs : Vec A n) →
                  tabulate (flip lookup xs) ≡ xs
tabulate∘lookup []       = refl
tabulate∘lookup (x ∷ xs) = P.cong (_∷_ x) $ tabulate∘lookup xs

-- If you look up an index in allFin n, then you get the index.

lookup-allFin : ∀ {n} (i : Fin n) → lookup i (allFin n) ≡ i
lookup-allFin = lookup∘tabulate id

-- Various lemmas relating lookup and _++_.

lookup-++-< : ∀ {a} {A : Set a} {m n}
              (xs : Vec A m) (ys : Vec A n) i (i<m : toℕ i < m) →
              lookup i (xs ++ ys) ≡ lookup (Fin.fromℕ≤ i<m) xs
lookup-++-< []       ys i       ()
lookup-++-< (x ∷ xs) ys zero    (s≤s z≤n)       = refl
lookup-++-< (x ∷ xs) ys (suc i) (s≤s (s≤s i<m)) =
  lookup-++-< xs ys i (s≤s i<m)

lookup-++-≥ : ∀ {a} {A : Set a} {m n}
              (xs : Vec A m) (ys : Vec A n) i (i≥m : toℕ i ≥ m) →
              lookup i (xs ++ ys) ≡ lookup (Fin.reduce≥ i i≥m) ys
lookup-++-≥ []       ys i       i≥m = refl
lookup-++-≥ (x ∷ xs) ys zero    ()
lookup-++-≥ (x ∷ xs) ys (suc i) (s≤s i≥m) = lookup-++-≥ xs ys i i≥m

lookup-++-inject+ : ∀ {a} {A : Set a} {m n}
                    (xs : Vec A m) (ys : Vec A n) i →
                    lookup (Fin.inject+ n i) (xs ++ ys) ≡ lookup i xs
lookup-++-inject+ []       ys ()
lookup-++-inject+ (x ∷ xs) ys zero    = refl
lookup-++-inject+ (x ∷ xs) ys (suc i) = lookup-++-inject+ xs ys i

lookup-++-+′ : ∀ {a} {A : Set a} {m n}
               (xs : Vec A m) (ys : Vec A n) i →
               lookup (fromℕ m +′ i) (xs ++ ys) ≡ lookup i ys
lookup-++-+′ []       ys       zero    = refl
lookup-++-+′ []       (y ∷ xs) (suc i) = lookup-++-+′ [] xs i
lookup-++-+′ (x ∷ xs) ys       i       = lookup-++-+′ xs ys i

-- map is a congruence.

map-cong : ∀ {a b n} {A : Set a} {B : Set b} {f g : A → B} →
           f ≗ g → _≗_ {A = Vec A n} (map f) (map g)
map-cong f≗g []       = refl
map-cong f≗g (x ∷ xs) = P.cong₂ _∷_ (f≗g x) (map-cong f≗g xs)

-- map is functorial.

map-id : ∀ {a n} {A : Set a} → _≗_ {A = Vec A n} (map id) id
map-id []       = refl
map-id (x ∷ xs) = P.cong (_∷_ x) (map-id xs)

map-∘ : ∀ {a b c n} {A : Set a} {B : Set b} {C : Set c}
        (f : B → C) (g : A → B) →
        _≗_ {A = Vec A n} (map (f ∘ g)) (map f ∘ map g)
map-∘ f g []       = refl
map-∘ f g (x ∷ xs) = P.cong (_∷_ (f (g x))) (map-∘ f g xs)

-- tabulate can be expressed using map and allFin.

tabulate-∘ : ∀ {n a b} {A : Set a} {B : Set b}
             (f : A → B) (g : Fin n → A) →
             tabulate (f ∘ g) ≡ map f (tabulate g)
tabulate-∘ {zero}  f g = refl
tabulate-∘ {suc n} f g =
  P.cong (_∷_ (f (g zero))) (tabulate-∘ f (g ∘ suc))

tabulate-allFin : ∀ {n a} {A : Set a} (f : Fin n → A) →
                  tabulate f ≡ map f (allFin n)
tabulate-allFin f = tabulate-∘ f id

-- The positive case of allFin can be expressed recursively using map.

allFin-map : ∀ n → allFin (suc n) ≡ zero ∷ map suc (allFin n)
allFin-map n = P.cong (_∷_ zero) $ tabulate-∘ suc id

-- If you look up every possible index, in increasing order, then you
-- get back the vector you started with.

map-lookup-allFin : ∀ {a} {A : Set a} {n} (xs : Vec A n) →
                    map (λ x → lookup x xs) (allFin n) ≡ xs
map-lookup-allFin {n = n} xs = begin
  map (λ x → lookup x xs) (allFin n) ≡⟨ P.sym $ tabulate-∘ (λ x → lookup x xs) id ⟩
  tabulate (λ x → lookup x xs)       ≡⟨ tabulate∘lookup xs ⟩
  xs                                 ∎
  where open P.≡-Reasoning

-- sum commutes with _++_.

sum-++-commute : ∀ {m n} (xs : Vec ℕ m) {ys : Vec ℕ n} →
                 sum (xs ++ ys) ≡ sum xs + sum ys
sum-++-commute []            = refl
sum-++-commute (x ∷ xs) {ys} = begin
  x + sum (xs ++ ys)
    ≡⟨ P.cong (λ p → x + p) (sum-++-commute xs) ⟩
  x + (sum xs + sum ys)
    ≡⟨ P.sym (+-assoc x (sum xs) (sum ys)) ⟩
  sum (x ∷ xs) + sum ys
    ∎
  where
  open P.≡-Reasoning
  open CommutativeSemiring Nat.commutativeSemiring hiding (_+_; sym)

-- foldr is a congruence.

foldr-cong : ∀ {a} {A : Set a}
               {b₁} {B₁ : ℕ → Set b₁}
               {f₁ : ∀ {n} → A → B₁ n → B₁ (suc n)} {e₁}
               {b₂} {B₂ : ℕ → Set b₂}
               {f₂ : ∀ {n} → A → B₂ n → B₂ (suc n)} {e₂} →
             (∀ {n x} {y₁ : B₁ n} {y₂ : B₂ n} →
                y₁ ≅ y₂ → f₁ x y₁ ≅ f₂ x y₂) →
             e₁ ≅ e₂ →
             ∀ {n} (xs : Vec A n) →
             foldr B₁ f₁ e₁ xs ≅ foldr B₂ f₂ e₂ xs
foldr-cong           _     e₁=e₂ []       = e₁=e₂
foldr-cong {B₁ = B₁} f₁=f₂ e₁=e₂ (x ∷ xs) =
  f₁=f₂ (foldr-cong {B₁ = B₁} f₁=f₂ e₁=e₂ xs)

-- foldl is a congruence.

foldl-cong : ∀ {a} {A : Set a}
               {b₁} {B₁ : ℕ → Set b₁}
               {f₁ : ∀ {n} → B₁ n → A → B₁ (suc n)} {e₁}
               {b₂} {B₂ : ℕ → Set b₂}
               {f₂ : ∀ {n} → B₂ n → A → B₂ (suc n)} {e₂} →
             (∀ {n x} {y₁ : B₁ n} {y₂ : B₂ n} →
                y₁ ≅ y₂ → f₁ y₁ x ≅ f₂ y₂ x) →
             e₁ ≅ e₂ →
             ∀ {n} (xs : Vec A n) →
             foldl B₁ f₁ e₁ xs ≅ foldl B₂ f₂ e₂ xs
foldl-cong           _     e₁=e₂ []       = e₁=e₂
foldl-cong {B₁ = B₁} f₁=f₂ e₁=e₂ (x ∷ xs) =
  foldl-cong {B₁ = B₁ ∘ suc} f₁=f₂ (f₁=f₂ e₁=e₂) xs

-- The (uniqueness part of the) universality property for foldr.

foldr-universal : ∀ {a b} {A : Set a} (B : ℕ → Set b)
                  (f : ∀ {n} → A → B n → B (suc n)) {e}
                  (h : ∀ {n} → Vec A n → B n) →
                  h [] ≡ e →
                  (∀ {n} x → h ∘ _∷_ x ≗ f {n} x ∘ h) →
                  ∀ {n} → h ≗ foldr B {n} f e
foldr-universal B f     h base step []       = base
foldr-universal B f {e} h base step (x ∷ xs) = begin
  h (x ∷ xs)
    ≡⟨ step x xs ⟩
  f x (h xs)
    ≡⟨ P.cong (f x) (foldr-universal B f h base step xs) ⟩
  f x (foldr B f e xs)
    ∎
  where open P.≡-Reasoning

-- A fusion law for foldr.

foldr-fusion : ∀ {a b c} {A : Set a}
                 {B : ℕ → Set b} {f : ∀ {n} → A → B n → B (suc n)} e
                 {C : ℕ → Set c} {g : ∀ {n} → A → C n → C (suc n)}
               (h : ∀ {n} → B n → C n) →
               (∀ {n} x → h ∘ f {n} x ≗ g x ∘ h) →
               ∀ {n} → h ∘ foldr B {n} f e ≗ foldr C g (h e)
foldr-fusion {B = B} {f} e {C} h fuse =
  foldr-universal C _ _ refl (λ x xs → fuse x (foldr B f e xs))

-- The identity function is a fold.

idIsFold : ∀ {a n} {A : Set a} → id ≗ foldr (Vec A) {n} _∷_ []
idIsFold = foldr-universal _ _ id refl (λ _ _ → refl)

-- Proof irrelevance for _[_]=_.

proof-irrelevance-[]= : ∀ {a} {A : Set a} {n} {xs : Vec A n} {i x} →
                        (p q : xs [ i ]= x) → p ≡ q
proof-irrelevance-[]= here            here             = refl
proof-irrelevance-[]= (there xs[i]=x) (there xs[i]=x') =
  P.cong there (proof-irrelevance-[]= xs[i]=x xs[i]=x')

-- _[_]=_ can be expressed using lookup and _≡_.

[]=↔lookup : ∀ {a n i} {A : Set a} {x} {xs : Vec A n} →
             xs [ i ]= x ↔ lookup i xs ≡ x
[]=↔lookup {i = i} {x = x} {xs} = record
  { to         = P.→-to-⟶ to
  ; from       = P.→-to-⟶ (from i xs)
  ; inverse-of = record
    { left-inverse-of  = λ _ → proof-irrelevance-[]= _ _
    ; right-inverse-of = λ _ → P.proof-irrelevance _ _
    }
  }
  where
  to : ∀ {n xs} {i : Fin n} → xs [ i ]= x → lookup i xs ≡ x
  to here            = refl
  to (there xs[i]=x) = to xs[i]=x

  from : ∀ {n} (i : Fin n) xs → lookup i xs ≡ x → xs [ i ]= x
  from zero    (.x ∷ _)  refl = here
  from (suc i) (_  ∷ xs) p    = there (from i xs p)

------------------------------------------------------------------------
-- Some properties related to _[_]≔_

[]≔-idempotent :
  ∀ {n a} {A : Set a} (xs : Vec A n) (i : Fin n) {x₁ x₂ : A} →
  (xs [ i ]≔ x₁) [ i ]≔ x₂ ≡ xs [ i ]≔ x₂
[]≔-idempotent []       ()
[]≔-idempotent (x ∷ xs) zero    = refl
[]≔-idempotent (x ∷ xs) (suc i) = P.cong (_∷_ x) $ []≔-idempotent xs i

[]≔-commutes :
  ∀ {n a} {A : Set a} (xs : Vec A n) (i j : Fin n) {x y : A} →
  i ≢ j → (xs [ i ]≔ x) [ j ]≔ y ≡ (xs [ j ]≔ y) [ i ]≔ x
[]≔-commutes []       ()      ()      _
[]≔-commutes (x ∷ xs) zero    zero    0≢0 = ⊥-elim $ 0≢0 refl
[]≔-commutes (x ∷ xs) zero    (suc i) _   = refl
[]≔-commutes (x ∷ xs) (suc i) zero    _   = refl
[]≔-commutes (x ∷ xs) (suc i) (suc j) i≢j =
  P.cong (_∷_ x) $ []≔-commutes xs i j (i≢j ∘ P.cong suc)

[]≔-updates : ∀ {n a} {A : Set a} (xs : Vec A n) (i : Fin n) {x : A} →
              (xs [ i ]≔ x) [ i ]= x
[]≔-updates []       ()
[]≔-updates (x ∷ xs) zero    = here
[]≔-updates (x ∷ xs) (suc i) = there ([]≔-updates xs i)

[]≔-minimal :
  ∀ {n a} {A : Set a} (xs : Vec A n) (i j : Fin n) {x y : A} →
  i ≢ j → xs [ i ]= x → (xs [ j ]≔ y) [ i ]= x
[]≔-minimal []       ()      ()      _   _
[]≔-minimal (x ∷ xs) .zero   zero    0≢0 here        = ⊥-elim $ 0≢0 refl
[]≔-minimal (x ∷ xs) .zero   (suc j) _   here        = here
[]≔-minimal (x ∷ xs) (suc i) zero    _   (there loc) = there loc
[]≔-minimal (x ∷ xs) (suc i) (suc j) i≢j (there loc) =
  there ([]≔-minimal xs i j (i≢j ∘ P.cong suc) loc)

map-[]≔ : ∀ {n a b} {A : Set a} {B : Set b}
          (f : A → B) (xs : Vec A n) (i : Fin n) {x : A} →
          map f (xs [ i ]≔ x) ≡ map f xs [ i ]≔ f x
map-[]≔ f []       ()
map-[]≔ f (x ∷ xs) zero    = refl
map-[]≔ f (x ∷ xs) (suc i) = P.cong (_∷_ _) $ map-[]≔ f xs i

[]≔-lookup : ∀ {a} {A : Set a} {n} (xs : Vec A n) (i : Fin n) →
             xs [ i ]≔ lookup i xs ≡ xs
[]≔-lookup []       ()
[]≔-lookup (x ∷ xs) zero    = refl
[]≔-lookup (x ∷ xs) (suc i) = P.cong (_∷_ x) $ []≔-lookup xs i