module Data.Int.Order where

Ordering integers🔗

The usual partial order on the integers relies on the observation that the number line looks like two copies of the natural numbers, smashed end-to-end at the number zero. This is precisely the definition of the order we use:

data _≤_ : Int → Int → Type where
  neg≤neg : ∀ {x y} → y Nat.≤ x → negsuc x ≤ negsuc y
  pos≤pos : ∀ {x y} → x Nat.≤ y → pos x    ≤ pos y
  neg≤pos : ∀ {x y}             → negsuc x ≤ pos y

Note the key properties: the ordering between negative numbers is reversed, and every negative number is smaller than every positive number. This means that $\mathbb{Z}$ decomposes, as an order, into an ordinal sum $\mathbb{N}^{\mathrm{op}} + \mathbb{N}$ .

Basic properties🔗

Proving that this is actually a partial order is a straightforward combination of case-bashing and appealing to the analogous properties for the ordering on natural numbers.

¬pos≤neg : ∀ {x y} → pos x ≤ negsuc y → ⊥
¬pos≤neg ()

≤-is-prop : ∀ {x y} → is-prop (x ≤ y)
≤-is-prop (neg≤neg p) (neg≤neg q) = ap neg≤neg (Nat.≤-is-prop p q)
≤-is-prop (pos≤pos p) (pos≤pos q) = ap pos≤pos (Nat.≤-is-prop p q)
≤-is-prop neg≤pos neg≤pos = refl

≤-refl : ∀ {x} → x ≤ x
≤-refl {x = pos x}    = pos≤pos Nat.≤-refl
≤-refl {x = negsuc x} = neg≤neg Nat.≤-refl

≤-trans : ∀ {x y z} → x ≤ y → y ≤ z → x ≤ z
≤-trans (neg≤neg p) (neg≤neg q) = neg≤neg (Nat.≤-trans q p)
≤-trans (neg≤neg p) neg≤pos     = neg≤pos
≤-trans (pos≤pos p) (pos≤pos q) = pos≤pos (Nat.≤-trans p q)
≤-trans neg≤pos (pos≤pos x)     = neg≤pos

≤-antisym : ∀ {x y} → x ≤ y → y ≤ x → x ≡ y
≤-antisym (neg≤neg p) (neg≤neg q) = ap negsuc (Nat.≤-antisym q p)
≤-antisym (pos≤pos p) (pos≤pos q) = ap pos (Nat.≤-antisym p q)

Totality🔗

The ordering on the integers is decidable, and moreover it is a total order. We show weak totality: if $x \not\le y$ , then $y \le x$ .

≤-is-weakly-total : ∀ x y → ¬ (x ≤ y) → y ≤ x
≤-is-weakly-total (pos    x) (pos    y) p = pos≤pos $
  Nat.≤-is-weakly-total x y (p ∘ pos≤pos)
≤-is-weakly-total (pos    x) (negsuc y) p = neg≤pos
≤-is-weakly-total (negsuc x) (pos    y) p = absurd (p neg≤pos)
≤-is-weakly-total (negsuc x) (negsuc y) p = neg≤neg $
  Nat.≤-is-weakly-total y x (p ∘ neg≤neg)

instance
  Dec-≤ : ∀ {x y} → Dec (x ≤ y)
  Dec-≤ {pos x} {pos y} with holds? (x Nat.≤ y)
  ... | yes p = yes (pos≤pos p)
  ... | no ¬p = no λ { (pos≤pos p) → ¬p p }
  Dec-≤ {negsuc x} {negsuc y} with holds? (y Nat.≤ x)
  ... | yes p = yes (neg≤neg p)
  ... | no ¬p = no λ { (neg≤neg p) → ¬p p }
  Dec-≤ {pos x} {negsuc y} = no ¬pos≤neg
  Dec-≤ {negsuc x} {pos y} = yes neg≤pos

Universal properties of maximum and minimum🔗

This case bash shows that maxℤ and minℤ satisfy their universal properties.

  maxℤ-≤l : (x y : Int) → x ≤ maxℤ x y
  maxℤ-≤l (pos x)    (pos y)    = pos≤pos (Nat.max-≤l x y)
  maxℤ-≤l (pos _)    (negsuc _) = ≤-refl
  maxℤ-≤l (negsuc _) (pos _)    = neg≤pos
  maxℤ-≤l (negsuc x) (negsuc y) = neg≤neg (Nat.min-≤l x y)

  maxℤ-≤r : (x y : Int) → y ≤ maxℤ x y
  maxℤ-≤r (pos x)    (pos y)    = pos≤pos (Nat.max-≤r x y)
  maxℤ-≤r (pos _)    (negsuc _) = neg≤pos
  maxℤ-≤r (negsuc _) (pos _)    = ≤-refl
  maxℤ-≤r (negsuc x) (negsuc y) = neg≤neg (Nat.min-≤r x y)

  maxℤ-univ : (x y z : Int) → x ≤ z → y ≤ z → maxℤ x y ≤ z
  maxℤ-univ _ _ _ (pos≤pos x≤z) (pos≤pos y≤z) = pos≤pos (Nat.max-univ _ _ _ x≤z y≤z)
  maxℤ-univ _ _ _ (pos≤pos x≤z) neg≤pos       = pos≤pos x≤z
  maxℤ-univ _ _ _ neg≤pos       (pos≤pos y≤z) = pos≤pos y≤z
  maxℤ-univ _ _ _ neg≤pos       neg≤pos       = neg≤pos
  maxℤ-univ _ _ _ (neg≤neg x≥z) (neg≤neg y≥z) = neg≤neg (Nat.min-univ _ _ _ x≥z y≥z)

  minℤ-≤l : (x y : Int) → minℤ x y ≤ x
  minℤ-≤l (pos x)    (pos y)    = pos≤pos (Nat.min-≤l x y)
  minℤ-≤l (pos _)    (negsuc _) = neg≤pos
  minℤ-≤l (negsuc _) (pos _)    = ≤-refl
  minℤ-≤l (negsuc x) (negsuc y) = neg≤neg (Nat.max-≤l x y)

  minℤ-≤r : (x y : Int) → minℤ x y ≤ y
  minℤ-≤r (pos x)    (pos y)    = pos≤pos (Nat.min-≤r x y)
  minℤ-≤r (pos _)    (negsuc _) = ≤-refl
  minℤ-≤r (negsuc _) (pos _)    = neg≤pos
  minℤ-≤r (negsuc x) (negsuc y) = neg≤neg (Nat.max-≤r x y)

  minℤ-univ : (x y z : Int) → z ≤ x → z ≤ y → z ≤ minℤ x y
  minℤ-univ _ _ _ (pos≤pos x≥z) (pos≤pos y≥z) = pos≤pos (Nat.min-univ _ _ _ x≥z y≥z)
  minℤ-univ _ _ _ neg≤pos       neg≤pos       = neg≤pos
  minℤ-univ _ _ _ neg≤pos       (neg≤neg y≤z) = neg≤neg y≤z
  minℤ-univ _ _ _ (neg≤neg x≤z) neg≤pos       = neg≤neg x≤z
  minℤ-univ _ _ _ (neg≤neg x≤z) (neg≤neg y≤z) = neg≤neg (Nat.max-univ _ _ _ x≤z y≤z)

Compatibility with the structure🔗

The last case bash in this module will establish that the ordering on integers is compatible with the successor, predecessor, and negation. Since addition is equivalent to iterated application of the successor and predecessor, we get as a corollary that addition also respects the order.

suc-≤ : ∀ x y → x ≤ y → sucℤ x ≤ sucℤ y
suc-≤ (pos x) (pos y) (pos≤pos p) = pos≤pos (Nat.s≤s p)
suc-≤ (negsuc zero) (pos y) p = pos≤pos Nat.0≤x
suc-≤ (negsuc zero) (negsuc zero) p = ≤-refl
suc-≤ (negsuc zero) (negsuc (suc y)) (neg≤neg ())
suc-≤ (negsuc (suc x)) (pos y) p = neg≤pos
suc-≤ (negsuc (suc x)) (negsuc zero) p = neg≤pos
suc-≤ (negsuc (suc x)) (negsuc (suc y)) (neg≤neg (Nat.s≤s p)) = neg≤neg p

pred-≤ : ∀ x y → x ≤ y → predℤ x ≤ predℤ y
pred-≤ posz posz p = ≤-refl
pred-≤ posz (possuc y) p = neg≤pos
pred-≤ (possuc x) posz (pos≤pos ())
pred-≤ (possuc x) (possuc y) (pos≤pos (Nat.s≤s p)) = pos≤pos p
pred-≤ (negsuc x) posz p = neg≤neg Nat.0≤x
pred-≤ (negsuc x) (possuc y) p = neg≤pos
pred-≤ (negsuc x) (negsuc y) (neg≤neg p) = neg≤neg (Nat.s≤s p)

rotℤ≤l : ∀ k x y → x ≤ y → rotℤ k x ≤ rotℤ k y
rotℤ≤l posz             x y p = p
rotℤ≤l (possuc k)       x y p = suc-≤ _ _ (rotℤ≤l (pos k) x y p)
rotℤ≤l (negsuc zero)    x y p = pred-≤ _ _ p
rotℤ≤l (negsuc (suc k)) x y p = pred-≤ _ _ (rotℤ≤l (negsuc k) x y p)

abstract
  +ℤ-mono-l : ∀ k x y → x ≤ y → (k +ℤ x) ≤ (k +ℤ y)
  +ℤ-mono-l k x y p = transport
    (sym (ap₂ _≤_ (rot-is-add k x) (rot-is-add k y)))
    (rotℤ≤l k x y p)

  +ℤ-mono-r : ∀ k x y → x ≤ y → (x +ℤ k) ≤ (y +ℤ k)
  +ℤ-mono-r k x y p = transport
    (ap₂ _≤_ (+ℤ-commutative k x) (+ℤ-commutative k y))
    (+ℤ-mono-l k x y p)

  negℤ-anti : ∀ x y → x ≤ y → negℤ y ≤ negℤ x
  negℤ-anti posz       posz       x≤y                     = x≤y
  negℤ-anti posz       (possuc y) _                       = neg≤pos
  negℤ-anti (possuc x) (possuc y) (pos≤pos (Nat.s≤s x≤y)) = neg≤neg x≤y
  negℤ-anti (negsuc _) posz       _                       = pos≤pos Nat.0≤x
  negℤ-anti (negsuc _) (possuc y) _                       = neg≤pos
  negℤ-anti (negsuc x) (negsuc y) (neg≤neg x≤y)           = pos≤pos (Nat.s≤s x≤y)

  negℤ-anti-full : ∀ x y → negℤ y ≤ negℤ x → x ≤ y
  negℤ-anti-full posz       (pos y)    _                       = pos≤pos Nat.0≤x
  negℤ-anti-full posz       (negsuc y) (pos≤pos ())
  negℤ-anti-full (possuc x) (possuc y) (neg≤neg x≤y)           = pos≤pos (Nat.s≤s x≤y)
  negℤ-anti-full (negsuc x) (pos y)    _                       = neg≤pos
  negℤ-anti-full (negsuc x) (negsuc y) (pos≤pos (Nat.s≤s y≤x)) = neg≤neg y≤x