Monotonicity and modalities for predicates over types with orders

Class/MonotonePredicate.agda

@@ -0,0 +1,3 @@
+module Class.MonotonePredicate where
+
+open import Class.MonotonePredicate.Core public 

Class/MonotonePredicate/Core.agda

@@ -0,0 +1,84 @@
+
+module Class.MonotonePredicate.Core where 
+
+open import Class.Prelude
+open import Class.HasOrder
+
+open import Relation.Unary
+open import Relation.Binary using (IsPreorder ; IsEquivalence)
+
+open import Level
+
+-- We make the "simplifying" assumption here that everything (carrier,
+-- relation, etc...) lives at the same level. Otherwise we couldn't
+-- express the (co)monadic structure of the possiblity and necessity
+-- modalities.
+module _ {A : Type ℓ} 
+         ⦃ _ : HasPreorder {A = A} {_≈_ = _≡_} {ℓ} {ℓ} ⦄ where
+
+  open IsPreorder {ℓ} ≤-isPreorder
+
+  record Monotone {ℓ′} (P : Pred A ℓ′) : Type (ℓ ⊔ ℓ′ ⊔ ℓ) where
+    field
+      weaken : ∀ {a a′} → a ≤ a′ → P a → P a′
+
+  record Antitone {ℓ′} (P : Pred A ℓ′) : Type (ℓ ⊔ ℓ′ ⊔ ℓ″) where
+    field
+      strengthen : ∀ {a a′} → a′ ≤ a → P a → P a′
+
+  open Monotone ⦃...⦄
+  open Antitone ⦃...⦄
+
+  -- The posibility modality. One way to think about posibility is as a unary
+  -- version of separating conjunction. 
+  record ◇ (P : Pred A ℓ) (a : A) : Set ℓ where
+    constructor ◇⟨_,_⟩ 
+    field
+      {a′}    : A
+      ι       : a′ ≤ a
+      px      : P a′ 
+
+  open ◇ public 
+
+  -- The necessity modality. In a similar spirit, we can view necessity as a unary
+  -- version of separating implication.
+  record □ (P : Pred A ℓ) (a : A) : Set ℓ where
+    constructor necessary 
+    field
+      □⟨_⟩_ : ∀ {a′} → (ι : a ≤ a′) → P a′
+
+  open □ public 
+
+  -- □ is a comonad over the category of monotone predicates over `A`  
+  extract : ∀ {P} → ∀[ □ P ⇒ P ]
+  extract px = □⟨ px ⟩ reflexive _≡_.refl
+
+  duplicate : ∀ {P} → ∀[ □ P ⇒ □ (□ P) ]
+  duplicate px = necessary λ ι → necessary λ ι′ → □⟨ px ⟩ trans ι ι′ 
+
+
+  -- ◇ is a monad over the category of monotone predicates over `A`.
+  return : ∀ {P} → ∀[ P ⇒ ◇ P ]
+  return px = ◇⟨ reflexive _≡_.refl , px ⟩
+
+  join : ∀ {P} → ∀[ ◇ (◇ P) ⇒ ◇ P ]
+  join ◇⟨ ι₁ , ◇⟨ ι₂ , px ⟩ ⟩ = ◇⟨ (trans ι₂ ι₁) , px ⟩
+
+  -- □ is right-adjoint to ◇
+  curry : ∀ {P Q} → ∀[ ◇ P ⇒ Q ] → ∀[ P ⇒ □ Q ]
+  curry f px = necessary (λ ι → f ◇⟨ ι , px ⟩)
+
+  uncurry : ∀ {P Q} → ∀[ P ⇒ □ Q ] → ∀[ ◇ P ⇒ Q ]
+  uncurry f ◇⟨ ι , px ⟩ = □⟨ f px ⟩ ι
+
+
+  -- The "Kripke exponent" (or, Kripke function space) between two predicates is
+  -- defined as the necessity of their implication.
+  _⇛_ : ∀ (P Q : Pred A ℓ) → Pred A ℓ 
+  P ⇛ Q = □ (P ⇒ Q) 
+
+  kripke-curry : {P Q R : Pred A ℓ} → ⦃ Monotone P ⦄ → ∀[ (P ∩ Q) ⇒ R ] → ∀[ P ⇒ (Q ⇛ R) ] 
+  kripke-curry f px₁ = necessary (λ i px₂ → f (weaken i px₁ , px₂))
+
+  kripke-uncurry : {P Q R : Pred A ℓ} → ∀[ P ⇒ (Q ⇛ R) ] → ∀[ (P ∩ Q) ⇒ R ] 
+  kripke-uncurry f (px₁ , px₂) = □⟨ f px₁ ⟩ reflexive _≡_.refl $ px₂