Library HoTT.Spaces.BinInt.Core

From HoTT Require Import Basics.
Require Import Spaces.Pos.Core.

Local Set Universe Minimization ToSet.

Binary Integers


Local Close Scope trunc_scope.
Local Close Scope nat_scope.

Local Open Scope positive_scope.

Definition of the Integers

We define an integer as being a positive number labelled negative, zero or a positive number labelled positive.
Inductive BinInt : Type0 :=
  | neg : Pos → BinInt
  | zero : BinInt
  | pos : Pos → BinInt.

Declare Scope binint_scope.
Local Open Scope binint_scope.
Delimit Scope binint_scope with binint.
Bind Scope binint_scope with BinInt.

The integers are a pointed type
Properties of constructors

Definition neg_inj {z w : Pos} (p : neg z = neg w) : z = w
  := transport (fun s ⇒ z =
    (match s with neg a ⇒ a | zero ⇒ w | pos a ⇒ w end)) p (idpath z).

Definition pos_inj {z w : Pos} (p : pos z = pos w) : z = w
  := transport (fun s ⇒ z =
    (match s with neg a ⇒ w | zero ⇒ w | pos a ⇒ a end)) p (idpath z).

Definition neg_neq_zero {z : Pos} : ¬ (neg z = zero)
  := fun p ⇒ transport (fun s ⇒
    match s with neg a ⇒ z = a| zero ⇒ Empty | pos _ ⇒ Empty end) p (idpath z).

Definition pos_neq_zero {z : Pos} : ¬ (pos z = zero)
  := fun p ⇒ transport (fun s ⇒
    match s with pos a ⇒ z = a | zero ⇒ Empty | neg _ ⇒ Empty end) p (idpath z).

Definition neg_neq_pos {z w : Pos} : ¬ (neg z = pos w)
  := fun p ⇒ transport (fun s ⇒
    match s with neg a ⇒ z = a | zero ⇒ Empty | pos _ ⇒ Empty end) p (idpath z).

Definition zero_neq_neg {z : Pos} := @neg_neq_zero z o symmetry _ _.

Definition zero_neq_pos {z : Pos} := @pos_neq_zero z o symmetry _ _.

Definition pos_neq_neg {z w : Pos} := @neg_neq_pos z w o symmetry _ _.

Conversion with a decimal representation for printing/parsing


Definition binint_to_decimal_binint (n : BinInt) : Decimal.int :=
  match n with
    | neg m ⇒ Decimal.Neg (pos_to_uint m)
    | zero ⇒ Decimal.Pos Decimal.Nil
    | pos m ⇒ Decimal.Pos (pos_to_uint m)
  end.

Definition binint_to_number_binint (n : BinInt) : Numeral.int :=
  Numeral.IntDec (binint_to_decimal_binint n).

Fixpoint binint_of_decimal_uint (d : Decimal.uint) : BinInt :=
  match d with
    | Decimal.Nil ⇒ zero
    | Decimal.D0 l ⇒ binint_of_decimal_uint l
    | Decimal.D1 l ⇒ pos (pos_of_uint_acc l 1)
    | Decimal.D2 l ⇒ pos (pos_of_uint_acc l 1~0)
    | Decimal.D3 l ⇒ pos (pos_of_uint_acc l 1~1)
    | Decimal.D4 l ⇒ pos (pos_of_uint_acc l 1~0~0)
    | Decimal.D5 l ⇒ pos (pos_of_uint_acc l 1~0~1)
    | Decimal.D6 l ⇒ pos (pos_of_uint_acc l 1~1~0)
    | Decimal.D7 l ⇒ pos (pos_of_uint_acc l 1~1~1)
    | Decimal.D8 l ⇒ pos (pos_of_uint_acc l 1~0~0~0)
    | Decimal.D9 l ⇒ pos (pos_of_uint_acc l 1~0~0~1)
  end.

Definition binint_of_decimal_binint (d : Decimal.int) : BinInt :=
  match d with
    | Decimal.Pos u ⇒ binint_of_decimal_uint u
    | Decimal.Neg u ⇒ let t := binint_of_decimal_uint u in
        match t with
          | pos v ⇒ neg v
          | _ ⇒ zero
        end
  end.

Definition binint_of_number_binint (d:Numeral.int) :=
  match d with
  | Numeral.IntDec d ⇒ Some (binint_of_decimal_binint d)
  | Numeral.IntHex _ ⇒ None
  end.

Number Notation BinInt binint_of_number_binint binint_to_number_binint : binint_scope.

(* For some reason 0 can be parsed as an integer, but is printed as zero. This notation fixes that. *)
Notation "0" := zero : binint_scope.

Doubling and variants


Definition binint_double x :=
  match x with
    | 0 ⇒ 0
    | pos p ⇒ pos p~0
    | neg p ⇒ neg p~0
  end.

Definition binint_succ_double x :=
  match x with
    | 0 ⇒ 1
    | pos p ⇒ pos p~1
    | neg p ⇒ neg (pos_pred_double p)
  end.

Definition binint_pred_double x :=
  match x with
    | 0 ⇒ neg 1
    | neg p ⇒ neg p~1
    | pos p ⇒ pos (pos_pred_double p)
  end.

Subtraction of positive into BinInt


Fixpoint binint_pos_sub (x y : Pos) {struct y} : BinInt :=
  match x, y with
    | p~1, q~1 ⇒ binint_double (binint_pos_sub p q)
    | p~1, q~0 ⇒ binint_succ_double (binint_pos_sub p q)
    | p~1, 1 ⇒ pos p~0
    | p~0, q~1 ⇒ binint_pred_double (binint_pos_sub p q)
    | p~0, q~0 ⇒ binint_double (binint_pos_sub p q)
    | p~0, 1 ⇒ pos (pos_pred_double p)
    | 1, q~1 ⇒ neg q~0
    | 1, q~0 ⇒ neg (pos_pred_double q)
    | 1, 1 ⇒ zero
  end%pos.

Negation


Definition binint_negation x :=
  match x with
    | zero ⇒ zero
    | pos x ⇒ neg x
    | neg x ⇒ pos x
  end.

Notation "- x" := (binint_negation x) : binint_scope.

Lemma ibnint_negation_negation n : --n = n.
Proof.
  by destruct n.
Qed.

Addition


Definition binint_add x y :=
  match x, y with
    | 0, y ⇒ y
    | x, 0 ⇒ x
    | pos x', pos y' ⇒ pos (x' + y')
    | pos x', neg y' ⇒ binint_pos_sub x' y'
    | neg x', pos y' ⇒ binint_pos_sub y' x'
    | neg x', neg y' ⇒ neg (x' + y')
  end.

Infix "+" := binint_add : binint_scope.

Successor


Definition binint_succ x := x + 1.

Predecessor


Definition binint_pred x := x + neg 1.

Subtraction


Definition binint_sub m n := m + -n.

Infix "-" := binint_sub : binint_scope.

Multiplication


Definition binint_mul x y :=
  match x, y with
    | 0, _ ⇒ 0
    | _, 0 ⇒ 0
    | pos x', pos y' ⇒ pos (x' × y')
    | pos x', neg y' ⇒ neg (x' × y')
    | neg x', pos y' ⇒ neg (x' × y')
    | neg x', neg y' ⇒ pos (x' × y')
  end.

Infix "×" := binint_mul : binint_scope.

Power function


Definition binint_pow x y :=
  match y with
    | pos p ⇒ pos_iter (binint_mul x) p 1
    | 0 ⇒ 1
    | neg _ ⇒ 0
  end.

Square


Definition binint_square x :=
  match x with
    | 0 ⇒ 0
    | pos p ⇒ pos (pos_square p)
    | neg p ⇒ pos (pos_square p)
  end.

Sign function


Definition binint_sgn z :=
  match z with
    | 0 ⇒ 0
    | pos p ⇒ 1
    | neg p ⇒ neg 1
  end.

(* ** Decidable paths and truncation. *)

Instance decpaths_binint : DecidablePaths BinInt.
Proof.
  intros [n | | n] [m | | m].
  + destruct (dec (n = m)) as [p | q].
    - apply inl, ap, p.
    - by apply inr; intro; apply q, neg_inj.
  + exact (inr neg_neq_zero).
  + exact (inr neg_neq_pos).
  + exact (inr zero_neq_neg).
  + exact (inl idpath).
  + exact (inr zero_neq_pos).
  + exact (inr pos_neq_neg).
  + exact (inr pos_neq_zero).
  + destruct (dec (n = m)) as [p | q].
    - apply inl, ap, p.
    - by apply inr; intro; apply q, pos_inj.
Defined.

Since integers have decidable paths they are a hset
Instance hset_binint : IsHSet BinInt | 0 := _.