I'm trying to prove that injective functions are left invertible in Coq. I've reached a point in my proof where my goal is an "exists" proposition. I want to define a function that uses terms from proof scope (types and functions I've intro'ed before) and then show the function to the "exists" goal. Here's what I wrote so far:
(* function composition *)
Definition fun_comp {A B C: Type} (f:A -> B) (g:B -> C) : A -> C :=
fun a: A => g (f a).
Notation "g .o f" := (fun_comp f g) (at level 70).
Definition nonempty (A: Type) := exists a: A, a = a.
(* identity function for any given type *)
Definition fun_id (A: Type) := fun a: A => a.
(* left invertible *)
Definition l_invertible {A B: Type} (f: A -> B) :=
exists fl:B->A, fl .o f = fun_id A.
Definition injective {A B: Type} (f: A -> B) :=
forall a a': A, f a = f a' -> a = a'.
(* is a given element in a function's image? *)
Definition elem_in_fun_image {A B: Type} (b: B) (f: A -> B) :=
exists a: A, f a = b.
Theorem injective_is_l_invertible:
forall (A B: Type) (f: A -> B), nonempty A /\ injective f -> l_invertible f.
Proof.
intros A B f H.
destruct H as [Hnempty Hinj].
unfold l_invertible.
unfold nonempty in Hnempty.
destruct Hnempty as [a0].
(* here would go my function definition and invoking "exists myfun" *)
Here's the function I'm trying to define:
Definition fL (b: B) := if elem_in_fun_image b f
then f a
else a0.
Here's what the proof window looks like:
1 subgoal
A : Type
B : Type
f : A -> B
a0 : A
H : a0 = a0
Hinj : injective f
========================= (1 / 1)
exists fl : B -> A, (fl .o f) = fun_id A
How do I do this? I'm very new to Coq so other comments and pointers are welcome.
This definition cannot be performed in the basic logic. You need to add in a few extra axioms:
(* from Coq.Logic.FunctionalExtensionality *)
functional_extensionality : forall A B (f g : A -> B),
(forall x, f x = g x) -> f = g
(* from Coq.Logic.Classical *)
classic : forall P : Prop, P \/ ~ P
(* from Coq.Logic.ClassicalChoice *)
choice : forall (A B : Type) (R : A->B->Prop),
(forall x : A, exists y : B, R x y) ->
exists f : A->B, (forall x : A, R x (f x)).
The goal is to define a relation R that characterizes the left inverse that you want to construct. The existentially quantified f will then be the inverse! You will need the classic axiom to show the precondition of choice, and you will need functional extensionality to show the equation that you want. I'll leave it as an exercise to find out what R needs to be and how to complete the proof.
Your script should start with the following line.
Require Import ClassicalChoice FunctionalEquality.
Because, as suggested by #arthur-azevedo-de-amorim, you will need these axioms.
Then, you should use choice with the relation "R y x" being
"f x = A or there is no element in A such whose image by f is y".
You will need the axiom classic to prove the existential statement that is required by choice:
assert (pointwise : forall y: B, exists x : A,
f x = y \/ (forall x : A f x <> y)).
choice will give you an existential statement for a function that returns the value you want. You only need to say that this function is the right one. You can give a name to that function by typing destruct (choice ... pointwise) (you have to fill in the ...).
You will have to prove an equality between two functions, but using the axiom functional_extensionality, you can reduce this problem to just proving that the two functions are equal on any x.
For that x, just instantiate the characteristic property of the function (as produced by destruct (choice ... pointwise) with the
value f x. There is a disjuction, but the right-hand side case is self-contradictory, because obviously f x is f x for some x.
For the left-hand side case, you will get an hypothesis of the form (I name the function produced by (choice ... pointwise) with the name it:
f (it (f x)) = f x
Here you can apply your injectivity assumption. to deduce that it (f x) = x.
This pretty much spells out the proof. In my own, experiment, I used classic, NNP, not_all_ex_not, functional_extensionality, which are lemmas coming from ClassicalChoice of FunctionalEquality.
Related
I tried to implement the following Coq code:
Set Implicit Arguments.
Inductive fun_eq A B (f : A -> B) : forall C D, (C -> D) -> Prop :=
fun_eqrefl : forall g : A -> B, f = g -> fun_eq f g.
Lemma fun_eq0 A B (f g : A -> B) : fun_eq f g -> f = g.
Proof. intros H. destruct H.
But destruct H. fails with an error message:
Abstracting over the terms "A", "B" and "g" leads to a term
fun (A0 B0 : Type) (g0 : A0 -> B0) => f = g0 which is ill-typed.
Reason is: Illegal application:
The term "#eq" of type "forall A : Type, A -> A -> Prop" cannot be applied to the terms
"A0 -> B0" : "Type"
"f" : "A -> B"
"g0" : "A0 -> B0"
The 2nd term has type "A -> B" which should be coercible to
"A0 -> B0".
I think there would be two workarounds for this kind of error, neither of which worked in this case.
One is to admit the proof_irrelevance. However, it is impossible to construct an alternating proof for fun_eq f g because the argument of fun_eqrefl is what we want.
Another way is to provide an exact term using refine, but I couldn't come up with such a term. Also, if there were such a term (it would involve match statement), I suspect my previous question could be solved in a similar way.
Is it possible to prove fun_eq0? If so, how can it be done?
You can prove it using UIP (special case of proof irrelevance for eq).
The trick is to rewrite one side of the equality f to cast eq_refl f, where cast : A = B -> A -> B, and use UIP/proof irrelevance to replace eq_refl with an equality proof obtained from the H : fun_eq f g assumption. That way, when you destruct H, the type of the LHS changes simultaneously with the type of the RHS.
Set Implicit Arguments.
From Coq Require Import ProofIrrelevance.
Inductive fun_eq A B (f : A -> B) : forall C D, (C -> D) -> Prop :=
fun_eqrefl : forall g : A -> B, f = g -> fun_eq f g.
(* Extract the equality on types *)
Definition fun_eq_tyeq {A B C D} (f : A -> B) (g : C -> D) (H : fun_eq f g) : (A -> B) = (C -> D) :=
match H with
| fun_eqrefl _ => eq_refl
end.
Definition cast A B (e : A = B) (x : A) : B := eq_rect A (fun T => T) x B e.
Lemma fun_eq0 A B (f g : A -> B) : fun_eq f g -> f = g.
Proof.
intros H.
change (cast eq_refl f = g).
replace eq_refl with (fun_eq_tyeq H) by apply UIP.
destruct H.
cbn.
auto.
Qed.
I am trying to prove that given
(eqX : relation X) (Hypo : Equivalence eqX) (f : X -> {x : X | P x})
then
eqX a b -> eqX (proj1_sig (f a)) (proj1_sig (f b))
The function f get a parameter of Type X and give an existing assertion {x : X | P x}. ( for example fun (n : nat) => {m : nat | S m = n} )
In one word, I would like to show that given two parameters which are equivalent under the equivalent relation eqX, then the destruct result of existing assertion {x : X | P x} is also of the same equivalence class.
Can I prove this goal directly(which means the Specif.sig hold this property), or I should prove or claim that f satisfy some constraint and after which can I get this assertion proven.
Your claim is not directly provable; consider
X := nat
eqX a b := (a mod 2) = (b mod 2)
P a := True
f x := exist P (x / 2) I
Then we have eqX 2 4 but we don't have eqX (proj1_sig (f 2)) (proj1_sig (f 4)) because we don't have eqX 1 2.
You can either take in the theorem you are trying to prove as a hypothesis, or you can take in a hypothesis of type forall a, proj1_sig (f a) = a, or you can take in a hypothesis of type forall a, eqX (proj1_sig (f a)) a. Note that all of these are provable (by reflexivity or intros; assumption) if you have f a := exist P a (g a) for some function g.
Is this what you are trying to show?
Require Import Coq.Relations.Relation_Definitions.
Require Import Coq.Classes.Equivalence.
Require Import Setoid.
Generalizable All Variables.
Lemma foo `{!#Equivalence A RA, #Equivalence B RB, f : #respecting A _ _ B _ _ , #equiv A _ _ a b} :
equiv (proj1_sig f a) (proj1_sig f b).
Proof.
now apply respecting_equiv.
Qed.
I'm working towards formalising Free Selective Applicative Functors in Coq, but struggling with proofs by induction for inductive data types with non-uniform type parameters.
Let me give a bit of an introduction on the datatype I'm dealing with.
In Haskell, we encode Free Selective Functors as a GADT:
data Select f a where
Pure :: a -> Select f a
Select :: Select f (Either a b) -> f (a -> b) -> Select f b
The crucial thing here is the existential type variable b in the second data constructor.
We can translate this definition to Coq:
Inductive Select (F : Type -> Type) (A : Set) : Set :=
Pure : A -> Select F A
| MkSelect : forall (B : Set), Select F (B + A) -> F (B -> A) -> Select F A.
As a side note, I use the -impredicative-set option to encode it.
Coq generates the following induction principle for this datatype:
Select_ind :
forall (F : Type -> Type) (P : forall A : Set, Select F A -> Prop),
(forall (A : Set) (a : A), P A (Pure a)) ->
(forall (A B : Set) (s : Select F (B + A)), P (B + A)%type s ->
forall f0 : F (B -> A), P A (MkSelect s f0)) ->
forall (A : Set) (s : Select F A), P A s
Here, the interesting bit is the predicate P : forall A : Set, Select F A -> Prop which is parametrised not only in the expression, but also in the expressions type parameter. As I understand, the induction principle has this particular form because of the first argument of the MkSelect constructor of type Select F (B + A).
Now, I would like to prove statements like the third Applicative law for the defined datatype:
Theorem Select_Applicative_law3
`{FunctorLaws F} :
forall (A B : Set) (u : Select F (A -> B)) (y : A),
u <*> pure y = pure (fun f => f y) <*> u.
Which involve values of type Select F (A -> B), i.e. expressions containing functions. However,
calling induction on variables of such types produces ill-typed terms. Consider an oversimplified example of an equality that can be trivially proved by reflexivity, but can't be proved using induction:
Lemma Select_induction_fail `{Functor F} :
forall (A B : Set) (a : A) (x : Select F (A -> B)),
Select_map (fun f => f a) x = Select_map (fun f => f a) x.
Proof.
induction x.
Coq complains with the error:
Error: Abstracting over the terms "P" and "x" leads to a term
fun (P0 : Set) (x0 : Select F P0) =>
Select_map (fun f : P0 => f a) x0 = Select_map (fun f : P0 => f a) x0
which is ill-typed.
Reason is: Illegal application (Non-functional construction):
The expression "f" of type "P0" cannot be applied to the term
"a" : "A"
Here, Coq can't construct the predicate abstracted over the type variable because the reversed function application from the statement becomes ill-typed.
My question is, how do I use induction on my datatype? I can't see a way how to modify the induction principle in such a way so the predicate would not abstract the type. I tried to use dependent induction, but it has been producing inductive hypothesis constrained by equalities similar to (A -> B -> C) = (X + (A -> B -> C)) which I think would not be possible to instantiate.
Please see the complete example on GitHub: https://github.com/tuura/selective-theory-coq/blob/impredicative-set/src/Control/Selective/RigidImpredSetMinimal.v
UPDATE:
Following the discussio in the gist I have tried to carry out proofs by induction on depth of expression. Unfortunately, this path was not very fruitful since the induction hypothesis I get in theorems similar to Select_Applicative_law3 appear to be unusable. I will leave this problem for now and will give it a try later.
Li-yao, many thanks again for helping me to improve my understanding!
Proofs by induction are motivated by recursive definitions. So to know what to apply induction to, look for Fixpoints.
Your Fixpoints most likely work on terms indexed by single type variables Select F A, that's exactly where you want to use induction, not at the toplevel of the goal.
A Fixpoint on terms indexed by function types A -> B is useless since no subterms of any Select term are indexed by function types. For the same reason, induction is useless on such terms.
Here I think the strong type discipline actually forces you to work everything out on paper before trying to do anything in Coq (which is a good thing in my opinion). Try to do the proof on paper, without even worrying about types; explicitly write down the predicate(s) you want to prove by induction. Here's a template to see what I mean:
By induction on u, we will show
u <*> pure x = pure (fun f => f x) <*> u
(* Dummy induction predicate for the sake of example. *)
(* Find the right one. *)
(* It may use quantifiers... *)
Base case (set u = Pure f). Prove:
Pure f <*> pure x = pure (fun f => f x) <*> Pure f
Induction step (set u = MkSelect v h). Prove:
MkSelect v h <*> pure x = pure (fun f => f x) <*> MkSelect v h
assuming the induction hypothesis for the subterm v (set u = v):
v <*> pure x = pure (fun f => f x) <*> v
Notice in particular that the last equation is ill-typed, but you can still run along with it to do equational reasoning. Regardless, it will likely turn out that there is no way to apply that hypothesis after simplifying the goal.
If you really need to use Coq to do some exploration, there is a trick, consisting in erasing the problematic type parameter (and all terms that depend on it). Depending on your familiarity with Coq, this tip may turn out to be more confusing than anything. So be careful.
The terms will still have the same recursive structure. Keep in mind that the proof should also follow the same structure, because the point is to add more types on top afterwards, so you should avoid shortcuts that rely on the lack of types if you can.
(* Replace all A and B by unit. *)
Inductive Select_ (F : unit -> Type) : Set :=
| Pure_ : unit -> Select_ F
| MkSelect_ : Select_ F -> F tt -> Select_ F
.
Arguments Pure_ {F}.
Arguments MkSelect_ {F}.
(* Example translating Select_map. The Functor f constraint gets replaced with a dummy function argument. *)
(* forall A B, (A -> B) -> (F A -> F B) *)
Fixpoint Select_map_ {F : unit -> Type} (fmap : forall t, unit -> (F t -> F t)) (f : unit -> unit) (v : Select_ F) : Select_ F :=
match v with
| Pure_ a => Pure_ (f a)
| MkSelect_ w h => MkSelect_ (Select_map_ fmap f w) (fmap _ tt h)
end.
With that, you can try to prove this trimmed down version of the functor laws for example:
Select_map_ fmap f (Select_map_ fmap g v) = Select_map_ fmap (fun x => f (g x)) v
(* Original theorem:
Select_map f (Select_map g v) = Select_map (fun x => f (g x)) v
*)
The point is that removing the parameter avoids the associated typing problems, so you can try to use induction naively to see how things (don't) work out.
Variables A B : Prop.
Theorem proj1 : A /\ B -> A.
In order to learn, I'm trying to prove this theorem by explicitly writing down a proof term using and_ind.
I would assume the correct proof term is
fun (H : A /\ B) => and_ind A B A (fun a _ => a) H
But this raises an error, and instead the correct term is
fun (H : A /\ B) => and_ind (fun a _ => a) H
I don't understand this. The definition of and_ind is
and_ind =
fun (A B P : Prop) (f : A -> B -> P) (a : A /\ B) => match a with
| conj x x0 => f x x0
end
: forall A B P : Prop, (A -> B -> P) -> A /\ B -> P
How can I see from that that the parameters (A B P : Prop) have to be omitted?
The "App" rule
from the Reference Manual seems to state clearly that quantified variables have to be explicitly "instantiated" using the function application syntax that I tried.
In Coq, you can declare some arguments of a function as implicit. When you call the function, you don't supply values for the implicit arguments; Coq automatically tries to infer suitable values, based on other information available during type checking. The A, B and P arguments of and_ind are all declared as implicit, and can be inferred from the type of the H argument and the result type of the function argument.
You can see what arguments are considered implicit with the About command:
About and_ind.
(* and_ind : forall A B P : Prop, (A -> B -> P) -> A /\ B -> P *)
(* Arguments A, B, P are implicit *)
(* Argument scopes are [type_scope type_scope type_scope function_scope _] *)
(* and_ind is transparent *)
(* Expands to: Constant Coq.Init.Logic.and_ind *)
You can turn off implicit arguments with an individual call with an # sign:
Check fun A B H => #and_ind A B A (fun a _ => a) H.
(* fun (A B : Prop) (H : A /\ B) => and_ind (fun (a : A) (_ : B) => a) H *)
(* : forall A B : Prop, A /\ B -> A *)
(Notice that Coq automatically omits implicit arguments when printing a term as well.)
The Coq manual has more information on that subject.
I try to formalize groups in Coq. I want to be as general as possible. I try to do something, but I'm not really happy with it. I found different implementations and I don't know which one to choose.
For example I found this :
https://people.cs.umass.edu/~arjun/courses/cs691pl-spring2014/assignments/groups.html
(* The set of the group. *)
Parameter G : Set.
(* The binary operator. *)
Parameter f : G -> G -> G.
(* The group identity. *)
Parameter e : G.
(* The inverse operator. *)
Parameter i : G -> G.
(* For readability, we use infix <+> to stand for the binary operator. *)
Infix "<+>" := f (at level 50, left associativity).
(* The operator [f] is associative. *)
Axiom assoc : forall a b c, a <+> b <+> c = a <+> (b <+> c).
(* [e] is the right-identity for all elements [a] *)
Axiom id_r : forall a, a <+> e = a.
(* [i a] is the right-inverse of [a]. *)
Axiom inv_r : forall a, a <+> i a = e.
but why the author use axioms and not definitions ? Moreover, I don't like to have some parameters at top level.
On the book CoqArt, I found this implementation :
Record group : Type :=
{A : Type;
op : A→A→A;
sym : A→A;
e : A;
e_neutral_left : ∀ x:A, op e x = x;
sym_op : ∀ x:A, op (sym x) x = e;
op_assoc : ∀ x y z:A, op (op x y) z = op x (op y z)}.
On this definition I think the definition is to specialize, because if I want to define monoïds, I will redefine op_assoc or neutre left. Beyond that,
for some theorems, I don't need to use groups. For example if I want to proove that right_inverse is the same as left_inverse if the law is associative.
An other question is what are good axioms for groups :
use neutral element as axiom or left neutral element
use inverse element as axiom or left inverse
What is the more convenient to work with ?
Finally, if I want to proove some other theorems, I probably want to have some syntactic sugar in order to use the binary operation and inverse elements. What are your advices to have a convenient notation for groups ?
For the moment I did this :
Definition binary_operation {S:Set} := S -> S -> S.
Definition commutative {S:Set} (dot:binary_operation) := forall (a b:S), dot a b = dot b a.
Definition associative {S:Set} (dot:binary_operation) := forall (a b c:S), dot (dot a b) c = dot a (dot b c).
Definition left_identity {S:Set} (dot:binary_operation) (e:S) := forall a:S, (dot e a) = a.
Definition right_identity {S:Set} (dot:binary_operation) (e:S) := forall a:S, (dot a e) = a.
Definition identity {S:Set} (dot: binary_operation) (e:S) := left_identity dot e /\ right_identity dot e.
Definition left_inv {S:Set} (dot:binary_operation) (a' a e:S) := identity dot e -> dot a' a = e.
Definition right_inv {S:Set} (dot:binary_operation) (a' a e:S) := identity dot e -> dot a a' = e.
Definition inv {S:Set} (dot:binary_operation) (a' a e:S) := left_inv dot a' a e /\ right_inv dot a' a e.
I found an implementation in the code source of Coq but I don't understand why it is a good implementation : https://github.com/tmiya/coq/blob/master/group/group2.v
I can't provide a complete answer, but perhaps this can help you a bit.
Your first article provides definitions and axioms sufficient for proving exercises without paying too much attention to being 'good' or 'useful' implementation. That's why axioms and not definitions.
If you want 'to be as general as possible', you can use example from CoqArt, or wrap your group definition in a section, using Variable instead of Parameter.
Section Group.
(* The set of the group. *)
Variable G : Set.
(* The binary operator. *)
Variable f : G -> G -> G.
(* The group identity. *)
Variable e : G.
(* The inverse operator. *)
Variable i : G -> G.
(* For readability, we use infix <+> to stand for the binary operator. *)
Infix "<+>" := f (at level 50, left associativity).
(* The operator [f] is associative. *)
Variable assoc : forall a b c, a <+> b <+> c = a <+> (b <+> c).
(* [e] is the right-identity for all elements [a] *)
Variable id_r : forall a, a <+> e = a.
(* [i a] is the right-inverse of [a]. *)
Variable inv_r : forall a, a <+> i a = e.
If you then prove some theorems inside this section, like this:
Theorem trivial : forall a b, a <+> e <+> b = a <+> b.
intros.
rewrite id_r.
auto.
Qed.
After section ends,
End Group.
Coq generalizes them
Check trivial.
trivial
: forall (G : Set) (f : G -> G -> G) (e : G),
(forall a : G, f a e = a) -> forall a b : G, f (f a e) b = f a b
As for your last example, it's probably not about actual definition of group, but instead about proving that a set, a binary operation and four axioms for this operation define a group.