{-# OPTIONS --without-K --safe #-} -- Definition of the untyped domain model -- -- This module defines variables definitions of the untyped domain model and its operations. module Mint.Semantics.Domain where open import Relation.Binary using (Rel; REL) open import Lib open import Mint.Statics.Syntax public mutual -- A local evaluation environment, which models an individual context -- -- It maps de Bruijn indices to a domain value. One should consider Env as a model -- of an infinitely long local substitution. It is convenient to not have to -- consider length restriction in an untyped setting and hence ease the -- formalization. In reality, we only rely on some finite prefix, as guaranteed by -- the completeness and soundness theorems. Env : Set Env = ℕ → D -- A (global) evaluation environment, which models a context stack -- -- It maps the context index to an associated local environment and the modal -- offset to the next local environment. -- -- The topmost(current) environment is at index 0. -- -- Similar to Env, an Envs models an infinitely long unified substitution. In -- addition to Env which keep tracks of the mapping from de Bruijn indices to values -- in each levels, we also keep track of modal offsets, which in unified -- substitutions keeps track of how many contexts to add to the codomain stack -- before going into the next world in Kripke semantics' sense. Envs : Set Envs = ℕ → ℕ × Env -- Domain values -- -- D models values in the untyped domain, both representing types and terms. -- -- Morally, D models β normal forms of Mint so it contains no β redices. To make D -- a "real" normal form, we must annotate it with another D representing the type -- value as done in Df. data D : Set where N : D □ : D → D Π : D → (t : Exp) → (ρ : Envs) → D -- t has 1 open var U : ℕ → D ze : D su : D → D Λ : (t : Exp) → (ρ : Envs) → D box : D → D ↑ : D → (c : Dn) → D -- Domain neutral values data Dn : Set where -- Variables in the domain model -- -- Notice that the number x here is not a de Bruijn index but an absolute -- representation of names. That is, this number does not change relative to the -- binding structure it currently exists in. l : (x : ℕ) → Dn rec : (T : Typ) → D → (t : Exp) → (ρ : Envs) → Dn → Dn -- T has 1 open var and t has 2 open vars _$_ : Dn → (d : Df) → Dn unbox : ℕ → Dn → Dn -- Domain normal values -- -- It is just D but we annotate it with another value representing its type, i.e. ↓ -- A b is a value b of type A. data Df : Set where ↓ : D → (a : D) → Df infixl 10 [_]_$′_ pattern l′ A x = ↑ A (l x) pattern unbox′ A n c = ↑ A (unbox n c) pattern [_]_$′_ A x y = ↑ A (_$_ x y) -- Untyped modal transformations -- -- A stream of natural numbers models an infinitely long modal transformation. All -- these numbers are modal offsets. UMoT : Set UMoT = ℕ → ℕ variable a a′ a″ : D b b′ b″ : D A A′ A″ : D B B′ B″ : D f f′ f″ : D c c′ c″ : Dn C C′ C″ : Dn d d′ d″ d‴ : Df ρ ρ′ ρ″ : Envs κ κ′ κ″ : UMoT -- Empty local evaluation environment -- -- Its right hand side is arbitrary as a well-typed term can never go out of bound. emp : Env emp n = ↑ N (l 0) -- Empty global evaluation environment -- -- An empty global evaluation environment can be considered as a substitution where no -- variable maps to a sensible value. empty : Envs empty n = 1 , emp infixl 4.3 _↦_ _↦′_ _↦′_ : Env → D → Env (ρ ↦′ d) zero = d (ρ ↦′ d) (suc x) = ρ x -- Push a value to the current local evaluation environment _↦_ : Envs → D → Envs (ρ ↦ d) 0 = proj₁ (ρ 0) , proj₂ (ρ 0) ↦′ d (ρ ↦ d) (suc n) = ρ (suc n) -- Grow the evaluation environment with a specified modal offset ext : Envs → ℕ → Envs ext ρ n zero = n , emp ext ρ n (suc m) = ρ m -- Truncation operation of evaluation environments C-Tr : Envs → ℕ → Envs C-Tr ρ n m = ρ (n + m) -- Drop the topmost value in the current local evaluation envrionment drop : Envs → Envs drop ρ zero = proj₁ (ρ 0) , λ m → proj₂ (ρ 0) (1 + m) drop ρ (suc n) = ρ (suc n) -- Look up the current local evaluation environment lookup : Envs → ℕ → D lookup ρ n = proj₂ (ρ 0) n -- Grow a UMoT by a modal offset ins : UMoT → ℕ → UMoT ins κ n zero = n ins κ n (suc m) = κ m -- Truncation of UMoTs instance UMoTHasTr : HasTr UMoT UMoTHasTr = record { _∥_ = λ κ n m → κ (n + m) } -- Truncation offsets of UMoTs M-O : UMoT → ℕ → ℕ M-O κ zero = 0 M-O κ (suc n) = κ 0 + M-O (κ ∥ 1) n instance MTransHasO : HasO UMoT MTransHasO = record { O = M-O } toUMoT : Envs → UMoT toUMoT ρ n = proj₁ (ρ n) instance -- Truncation offsets of evaluation environments are essentially the same EnvsHasO : HasO Envs EnvsHasO = record { O = λ ρ → M-O (toUMoT ρ) } EnvHasTr : HasTr Envs EnvHasTr = record { _∥_ = C-Tr } -- Composition of UMoTs infixl 3 _ø_ _ø_ : UMoT → UMoT → UMoT (κ ø κ′) zero = O κ′ (κ 0) (κ ø κ′) (suc n) = (κ ∥ 1 ø κ′ ∥ κ 0) n -- Monotonicity of the domain model / Application of UMoTs to the domain model mutual mtran : D → UMoT → D mtran N κ = N mtran (□ A) κ = □ (mtran A (ins κ 1)) mtran (Π A t ρ) κ = Π (mtran A κ) t (mtran-Envs ρ κ) mtran (U i) κ = U i mtran ze κ = ze mtran (su a) κ = su (mtran a κ) mtran (Λ t ρ) κ = Λ t (mtran-Envs ρ κ) mtran (box a) κ = box (mtran a (ins κ 1)) mtran (↑ A c) κ = ↑ (mtran A κ) (mtran-c c κ) mtran-c : Dn → UMoT → Dn mtran-c (l x) κ = l x mtran-c (rec T a t ρ c) κ = rec T (mtran a κ) t (mtran-Envs ρ κ) (mtran-c c κ) mtran-c (c $ d) κ = mtran-c c κ $ mtran-d d κ mtran-c (unbox n c) κ = unbox (O κ n) (mtran-c c (κ ∥ n)) mtran-d : Df → UMoT → Df mtran-d (↓ A a) κ = ↓ (mtran A κ) (mtran a κ) mtran-Envs : Envs → UMoT → Envs mtran-Envs ρ κ n = O (κ ∥ O ρ n) (proj₁ (ρ n)) , λ m → mtran (proj₂ (ρ n) m) (κ ∥ O ρ n) instance DMonotone : Monotone D UMoT DMonotone = record { _[_] = mtran } DnMonotone : Monotone Dn UMoT DnMonotone = record { _[_] = mtran-c } DfMonotone : Monotone Df UMoT DfMonotone = record { _[_] = mtran-d } EnvsMonotone : Monotone Envs UMoT EnvsMonotone = record { _[_] = mtran-Envs } -- Identity UMoT vone : UMoT vone _ = 1 -- Relational syntax infix 1 _≈_∈_ _∈′_ _∼_∈_ _≈_∈_ : ∀ {i} {A : Set i} → A → A → Rel A i → Set i a ≈ b ∈ P = P a b _∈′_ : ∀ {i} {A : Set i} → A → Rel A i → Set i a ∈′ P = P a a _∼_∈_ : ∀ {i} {A B : Set i} → A → B → REL A B i → Set i a ∼ b ∈ P = P a b