-- Copyright (c) 2014 Microsoft Corporation. All rights reserved. -- Released under Apache 2.0 license as described in the file LICENSE. -- Author: Jeremy Avigad, Jakob von Raumer -- Ported from Coq HoTT import .path open path function -- Equivalences -- ------------ definition Sect {A B : Type} (s : A → B) (r : B → A) := Πx : A, r (s x) ≈ x -- -- TODO: need better means of declaring structures -- -- TODO: note that Coq allows projections to be declared to be coercions on the fly -- Structure IsEquiv inductive IsEquiv [class] {A B : Type} (f : A → B) := mk : Π (inv : B → A) (retr : Sect inv f) (sect : Sect f inv) (adj : Πx, retr (f x) ≈ ap f (sect x)), IsEquiv f namespace IsEquiv definition inv {A B : Type} (f : A → B) [H : IsEquiv f] : B → A := IsEquiv.rec (λinv retr sect adj, inv) H -- TODO: note: does not type check without giving the type definition retr {A B : Type} (f : A → B) [H : IsEquiv f] : Sect (inv f) f := IsEquiv.rec (λinv retr sect adj, retr) H definition sect {A B : Type} (f : A → B) [H : IsEquiv f] : Sect f (inv f) := IsEquiv.rec (λinv retr sect adj, sect) H definition adj {A B : Type} (f : A → B) [H : IsEquiv f] : Πx, retr f (f x) ≈ ap f (sect f x) := IsEquiv.rec (λinv retr sect adj, adj) H postfix `⁻¹` := inv end IsEquiv -- Structure Equiv inductive Equiv (A B : Type) : Type := mk : Π (equiv_fun : A → B) (equiv_isequiv : IsEquiv equiv_fun), Equiv A B namespace Equiv --Note: No coercion here definition equiv_fun {A B : Type} (e : Equiv A B) : A → B := Equiv.rec (λequiv_fun equiv_isequiv, equiv_fun) e definition equiv_isequiv [instance] {A B : Type} (e : Equiv A B) : IsEquiv (equiv_fun e) := Equiv.rec (λequiv_fun equiv_isequiv, equiv_isequiv) e infix `≃`:25 := Equiv end Equiv -- Some instances and closure properties of equivalences namespace IsEquiv variables {A B C : Type} {f : A → B} {g : B → C} {f' : A → B} -- The identity function is an equivalence. definition id_closed [instance] : (@IsEquiv A A id) := IsEquiv.mk id (λa, idp) (λa, idp) (λa, idp) -- The composition of two equivalences is, again, an equivalence. definition comp_closed [instance] (Hf : IsEquiv f) (Hg : IsEquiv g) : (IsEquiv (g ∘ f)) := IsEquiv.mk ((inv f) ∘ (inv g)) (λc, ap g (retr f (g⁻¹ c)) ⬝ retr g c) (λa, ap (inv f) (sect g (f a)) ⬝ sect f a) (λa, (whiskerL _ (adj g (f a))) ⬝ (ap_pp g _ _)⁻¹ ⬝ ap02 g (concat_A1p (retr f) (sect g (f a))⁻¹ ⬝ (ap_compose (inv f) f _ ◾ adj f a) ⬝ (ap_pp f _ _)⁻¹ ) ⬝ (ap_compose f g _)⁻¹ ) -- Any function equal to an equivalence is an equivlance as well. definition path_closed (Hf : IsEquiv f) (Heq : f ≈ f') : (IsEquiv f') := path.rec_on Heq Hf -- Any function pointwise equal to an equivalence is an equivalence as well. definition homotopic (Hf : IsEquiv f) (Heq : f ∼ f') : (IsEquiv f') := let sect' := (λ b, (Heq (inv f b))⁻¹ ⬝ retr f b) in let retr' := (λ a, (ap (inv f) (Heq a))⁻¹ ⬝ sect f a) in let adj' := (λ (a : A), let ff'a := Heq a in let invf := inv f in let secta := sect f a in let retrfa := retr f (f a) in let retrf'a := retr f (f' a) in have eq1 : _ ≈ _, from calc ap f secta ⬝ ff'a ≈ retrfa ⬝ ff'a : (ap _ (adj f _ ))⁻¹ ... ≈ ap (f ∘ invf) ff'a ⬝ retrf'a : !concat_A1p⁻¹ ... ≈ ap f (ap invf ff'a) ⬝ retr f (f' a) : {ap_compose invf f _}, have eq2 : _ ≈ _, from calc retrf'a ≈ (ap f (ap invf ff'a))⁻¹ ⬝ (ap f secta ⬝ ff'a) : moveL_Vp _ _ _ (eq1⁻¹) ... ≈ ap f (ap invf ff'a)⁻¹ ⬝ (ap f secta ⬝ Heq a) : {ap_V invf ff'a} ... ≈ ap f (ap invf ff'a)⁻¹ ⬝ (Heq (invf (f a)) ⬝ ap f' secta) : {!concat_Ap} ... ≈ (ap f (ap invf ff'a)⁻¹ ⬝ Heq (invf (f a))) ⬝ ap f' secta : {!concat_pp_p⁻¹} ... ≈ (ap f ((ap invf ff'a)⁻¹) ⬝ Heq (invf (f a))) ⬝ ap f' secta : {!ap_V⁻¹} ... ≈ (Heq (invf (f' a)) ⬝ ap f' ((ap invf ff'a)⁻¹)) ⬝ ap f' secta : {!concat_Ap} ... ≈ (Heq (invf (f' a)) ⬝ (ap f' (ap invf ff'a))⁻¹) ⬝ ap f' secta : {!ap_V} ... ≈ Heq (invf (f' a)) ⬝ ((ap f' (ap invf ff'a))⁻¹ ⬝ ap f' secta) : !concat_pp_p, have eq3 : _ ≈ _, from calc (Heq (invf (f' a)))⁻¹ ⬝ retr f (f' a) ≈ (ap f' (ap invf ff'a))⁻¹ ⬝ ap f' secta : moveR_Vp _ _ _ eq2 ... ≈ (ap f' ((ap invf ff'a)⁻¹)) ⬝ ap f' secta : {!ap_V⁻¹} ... ≈ ap f' ((ap invf ff'a)⁻¹ ⬝ secta) : !ap_pp⁻¹, eq3) in IsEquiv.mk (inv f) sect' retr' adj' end IsEquiv namespace IsEquiv variables {A B : Type} (f : A → B) (g : B → A) (ret : Sect g f) (sec : Sect f g) context set_option unifier.max_steps 30000 --To construct an equivalence it suffices to state the proof that the inverse is a quasi-inverse. definition adjointify : IsEquiv f := let sect' := (λx, ap g (ap f (inverse (sec x))) ⬝ ap g (ret (f x)) ⬝ sec x) in let adj' := (λ (a : A), let fgretrfa := ap f (ap g (ret (f a))) in let fgfinvsect := ap f (ap g (ap f ((sec a)⁻¹))) in let fgfa := f (g (f a)) in let retrfa := ret (f a) in have eq1 : ap f (sec a) ≈ _, from calc ap f (sec a) ≈ idp ⬝ ap f (sec a) : !concat_1p⁻¹ ... ≈ (ret (f a) ⬝ (ret (f a)⁻¹)) ⬝ ap f (sec a) : {!concat_pV⁻¹} ... ≈ ((ret (fgfa))⁻¹ ⬝ ap (f ∘ g) (ret (f a))) ⬝ ap f (sec a) : {!concat_pA1⁻¹} ... ≈ ((ret (fgfa))⁻¹ ⬝ fgretrfa) ⬝ ap f (sec a) : {ap_compose g f _} ... ≈ (ret (fgfa))⁻¹ ⬝ (fgretrfa ⬝ ap f (sec a)) : !concat_pp_p, have eq2 : ap f (sec a) ⬝ idp ≈ (ret fgfa)⁻¹ ⬝ (fgretrfa ⬝ ap f (sec a)), from !concat_p1 ⬝ eq1, have eq3 : idp ≈ _, from calc idp ≈ (ap f (sec a))⁻¹ ⬝ ((ret fgfa)⁻¹ ⬝ (fgretrfa ⬝ ap f (sec a))) : moveL_Vp _ _ _ eq2 ... ≈ (ap f (sec a)⁻¹ ⬝ (ret fgfa)⁻¹) ⬝ (fgretrfa ⬝ ap f (sec a)) : !concat_p_pp ... ≈ (ap f ((sec a)⁻¹) ⬝ (ret fgfa)⁻¹) ⬝ (fgretrfa ⬝ ap f (sec a)) : {!ap_V⁻¹} ... ≈ ((ap f ((sec a)⁻¹) ⬝ (ret fgfa)⁻¹) ⬝ fgretrfa) ⬝ ap f (sec a) : !concat_p_pp ... ≈ ((retrfa⁻¹ ⬝ ap (f ∘ g) (ap f ((sec a)⁻¹))) ⬝ fgretrfa) ⬝ ap f (sec a) : {!concat_pA1⁻¹} ... ≈ ((retrfa⁻¹ ⬝ fgfinvsect) ⬝ fgretrfa) ⬝ ap f (sec a) : {ap_compose g f _} ... ≈ (retrfa⁻¹ ⬝ (fgfinvsect ⬝ fgretrfa)) ⬝ ap f (sec a) : {!concat_p_pp⁻¹} ... ≈ retrfa⁻¹ ⬝ ap f (ap g (ap f ((sec a)⁻¹)) ⬝ ap g (ret (f a))) ⬝ ap f (sec a) : {!ap_pp⁻¹} ... ≈ retrfa⁻¹ ⬝ (ap f (ap g (ap f ((sec a)⁻¹)) ⬝ ap g (ret (f a))) ⬝ ap f (sec a)) : !concat_p_pp⁻¹ ... ≈ retrfa⁻¹ ⬝ ap f ((ap g (ap f ((sec a)⁻¹)) ⬝ ap g (ret (f a))) ⬝ sec a) : {!ap_pp⁻¹}, have eq4 : ret (f a) ≈ ap f ((ap g (ap f ((sec a)⁻¹)) ⬝ ap g (ret (f a))) ⬝ sec a), from moveR_M1 _ _ eq3, eq4) in IsEquiv.mk g ret sect' adj' end end IsEquiv namespace IsEquiv variables {A B: Type} (f : A → B) --The inverse of an equivalence is, again, an equivalence. definition inv_closed [instance] [Hf : IsEquiv f] : (IsEquiv (inv f)) := adjointify (inv f) f (sect f) (retr f) end IsEquiv namespace IsEquiv variables {A : Type} section variables {B C : Type} (f : A → B) {f' : A → B} [Hf : IsEquiv f] include Hf definition cancel_R (g : B → C) [Hgf : IsEquiv (g ∘ f)] : (IsEquiv g) := homotopic (comp_closed !inv_closed Hgf) (λb, ap g (retr f b)) definition cancel_L (g : C → A) [Hgf : IsEquiv (f ∘ g)] : (IsEquiv g) := homotopic (comp_closed Hgf !inv_closed) (λa, sect f (g a)) --Rewrite rules definition moveR_M {x : A} {y : B} (p : x ≈ (inv f) y) : (f x ≈ y) := (ap f p) ⬝ (retr f y) definition moveL_M {x : A} {y : B} (p : (inv f) y ≈ x) : (y ≈ f x) := (moveR_M f (p⁻¹))⁻¹ definition moveR_V {x : B} {y : A} (p : x ≈ f y) : (inv f) x ≈ y := ap (inv f) p ⬝ sect f y definition moveL_V {x : B} {y : A} (p : f y ≈ x) : y ≈ (inv f) x := (moveR_V f (p⁻¹))⁻¹ definition ap_closed [instance] (x y : A) : IsEquiv (ap f) := adjointify (ap f) (λq, (inverse (sect f x)) ⬝ ap (f⁻¹) q ⬝ sect f y) (λq, !ap_pp ⬝ whiskerR !ap_pp _ ⬝ ((!ap_V ⬝ inverse2 ((adj f _)⁻¹)) ◾ (inverse (ap_compose (f⁻¹) f _)) ◾ (adj f _)⁻¹) ⬝ concat_pA1_p (retr f) _ _ ⬝ whiskerR !concat_Vp _ ⬝ !concat_1p) (λp, whiskerR (whiskerL _ ((ap_compose f (f⁻¹) _)⁻¹)) _ ⬝ concat_pA1_p (sect f) _ _ ⬝ whiskerR !concat_Vp _ ⬝ !concat_1p) -- The function equiv_rect says that given an equivalence f : A → B, -- and a hypothesis from B, one may always assume that the hypothesis -- is in the image of e. -- In fibrational terms, if we have a fibration over B which has a section -- once pulled back along an equivalence f : A → B, then it has a section -- over all of B. definition equiv_rect (P : B -> Type) : (Πx, P (f x)) → (Πy, P y) := (λg y, path.transport _ (retr f y) (g (f⁻¹ y))) definition equiv_rect_comp (P : B → Type) (df : Π (x : A), P (f x)) (x : A) : equiv_rect f P df (f x) ≈ df x := let eq1 := (apD df (sect f x)) in calc equiv_rect f P df (f x) ≈ transport P (retr f (f x)) (df (f⁻¹ (f x))) : idp ... ≈ transport P (ap f (sect f x)) (df (f⁻¹ (f x))) : adj f ... ≈ transport (P ∘ f) (sect f x) (df (f⁻¹ (f x))) : transport_compose ... ≈ df x : eq1 end --Transporting is an equivalence protected definition transport [instance] (P : A → Type) {x y : A} (p : x ≈ y) : (IsEquiv (transport P p)) := IsEquiv.mk (transport P (p⁻¹)) (transport_pV P p) (transport_Vp P p) (transport_pVp P p) end IsEquiv namespace Equiv context parameters {A B C : Type} (eqf : A ≃ B) private definition f : A → B := equiv_fun eqf private definition Hf : IsEquiv f := equiv_isequiv eqf protected definition id : A ≃ A := Equiv.mk id IsEquiv.id_closed protected theorem compose (eqg: B ≃ C) : A ≃ C := Equiv.mk ((equiv_fun eqg) ∘ f) (IsEquiv.comp_closed Hf (equiv_isequiv eqg)) theorem path_closed (f' : A → B) (Heq : equiv_fun eqf ≈ f') : A ≃ B := Equiv.mk f' (IsEquiv.path_closed Hf Heq) theorem inv_closed : B ≃ A := Equiv.mk (IsEquiv.inv f) !IsEquiv.inv_closed theorem cancel_R {g : B → C} (Hgf : IsEquiv (g ∘ f)) : B ≃ C := Equiv.mk g (IsEquiv.cancel_R f _) theorem cancel_L {g : C → A} (Hgf : IsEquiv (f ∘ g)) : C ≃ A := Equiv.mk g (IsEquiv.cancel_L f _) protected theorem transport (P : A → Type) {x y : A} {p : x ≈ y} : (P x) ≃ (P y) := Equiv.mk (transport P p) (IsEquiv.transport P p) end context parameters {A B : Type} (eqf eqg : A ≃ B) private definition Hf [instance] : IsEquiv (equiv_fun eqf) := equiv_isequiv eqf private definition Hg [instance] : IsEquiv (equiv_fun eqg) := equiv_isequiv eqg theorem inv_eq (p : eqf ≈ eqg) : IsEquiv.inv (equiv_fun eqf) ≈ IsEquiv.inv (equiv_fun eqg) := path.rec_on p idp end -- calc enviroment -- Note: Calculating with substitutions needs univalence calc_trans compose calc_refl id calc_symm inv_closed end Equiv