2

In Coq, I showed the associativity of append on vectors using:

Require Import Coq.Vectors.VectorDef Omega.
Program Definition t_app_assoc v p q r (a : t v p) (b : t v q) (c : t v r) :=
  append (append a b) c = append a (append b c).
Next Obligation. omega. Qed.

I now want to apply this equality in a proof. Below is the easiest goal that I would expect to be provable with t_app_assoc. Of course it can be proven by simpl - this is just an example.

Goal (append (append (nil nat) (nil _)) (nil _)
   = append (nil _) (append (nil _) (nil _))).
apply t_app_assoc.

This apply fails with:

Error: Unable to unify "Prop" with
 "append (append (nil nat) (nil nat)) (nil nat) =
  append (nil nat) (append (nil nat) (nil nat))".

How can I apply t_app_assoc? Or is there a better way to define it? I thought I needed a Program Definition, because simply using a Lemma leads to a type error because t v (p + (q + r)) and t v (p + q + r) are not the same to Coq.

1 Answers1

2

Prologue

I guess what you want to to is to prove that the vector concatenation is associative and then use that fact as a lemma.

But t_app_assoc as you define it has the following type:

t_app_assoc
     : forall (v : Type) (p q r : nat), t v p -> t v q -> t v r -> Prop

You basically want to use : instead of := as follows.

From Coq Require Import Vector Arith.
Import VectorNotations.
Import EqNotations.  (* rew notation, see below *)

Section Append.

Variable A : Type.
Variable p q r : nat.
Variables (a : t A p) (b : t A q) (c : t A r).

Fail Lemma t_app_assoc :
  append (append a b) c = append a (append b c).

Unfortunately, we cannot even state a lemma like this using the usual homogeneous equality.

The left-hand side has the following type:

Check append (append a b) c : t A (p + q + r).

whereas the right-hand side is of type

Check append a (append b c) : t A (p + (q + r)).

Since t A (p + q + r) is not the same as t A (p + (q + r)) we cannot use = to state the above lemma.

Let me describe some ways of working around this issue:

rew notation

Lemma t_app_assoc_rew :
  append (append a b) c = rew (plus_assoc _ _ _) in
  append a (append b c).
Admitted.

Here we just use the law of associativity of addition for natural numbers to cast the type of RHS to t A (p + q + r).

To make it work one needs to Import EqNotations. before.

cast function

This is a common problem, so the authors of the Vector library decided to provide a cast function with the following type:

cast :
  forall (A : Type) (m : nat),
  Vector.t A m -> forall n : nat, m = n -> Vector.t A n

Let me show how one can use it to prove the law of associativity for vectors. But let's prove the following auxiliary lemma first:

Lemma uncast {X n} {v : Vector.t X n} e :
  cast v e = v.
Proof. induction v as [|??? IH]; simpl; rewrite ?IH; reflexivity. Qed.

Now we are all set:

Lemma t_app_assoc_cast (a : t A p) (b : t A q) (c : t A r) :
  append (append a b) c = cast (append a (append b c)) (plus_assoc _ _ _).
Proof.
  generalize (Nat.add_assoc p q r).
  induction a as [|h p' a' IH]; intros e.
  - now rewrite uncast.
  - simpl; f_equal. apply IH.
Qed.

Heterogeneous equality (a.k.a. John Major equality)

Lemma t_app_assoc_jmeq :
  append (append a b) c ~= append a (append b c).
Admitted.

End Append.

If you compare the definition of the homogeneous equality

Inductive eq (A : Type) (x : A) : A -> Prop :=
  eq_refl : x = x.

and the definition of heterogeneous equality

Inductive JMeq (A : Type) (x : A) : forall B : Type, B -> Prop :=
  JMeq_refl : x ~= x.

you will see that with JMeq the LHS and RHS don't have to be of the same type and this is why the statement of t_app_assoc_jmeq looks a bit simpler than the previous ones.

Other approaches to vectors

See e.g. this question and this one; I also find this answer very useful too.

Anton Trunov
  • 15,074
  • 2
  • 23
  • 43
  • This has been very helpful, thanks! Only, I haven't been able to prove the rew/cast lemma (nor found anyone using it...). I actually have no idea where to start; unfold cast does not lead to anything useful. –  May 14 '18 at 14:11