Polynomials and a reflexive tactic for solving boolean goals (using heyting or classical boolean algebra). Original version by Rafael Bocquet, 2016. Updated to use Equations for all definitions by M. Sozeau, 2016-2017. If running this interactively you can ignore the printing and hide directives which are just used to instruct coqdoc.

We start with a simple definition deciding if some integer is equal to 0 or not. Integers are encoded using an inductive type Z with three constructors Z0, Zpos and Zneg, the latter two taking positive numbers as arguments. There is a single representant of 0 which we discriminate here. The second clause actually captures both the Zpos and Zneg constructors.

Equations IsNZ (z : Z) : bool :=
  IsNZ Z0 := false; IsNZ _ := true.

The specification of this test is that it returns true iff the variable is indeed different from 0 w.r.t. the standard Leibniz equality. We elide a simple proof by case analysis. Note that we use an implicit coercion from bool to Prop here, as is usual when doing boolean reflection.

Lemma IsNZ_spec z : IsNZ z (z 0)%Z.
  funelim (IsNZ z); unfold not; split; intros;
    (discriminate || contradiction || constructor).

Multivariate polynomials

Using an indexed inductive type, we ensure that polynomials of have a unique representation. The first index indicates that the polynom is null. The second index gives the number of free variables.

Inductive poly : bool nat Type :=
| poly_z : poly true O
| poly_c (z : Z) : IsNZ z poly false O
| poly_l {n b} (Q : poly b n) : poly b (S n)
| poly_s {n b} (P : poly b n) (Q : poly false (S n)) :
    poly false (S n).

  • poly_z represents the null polynomial.
  • poly_c c represents the constant polynomial c where c is non-zero (i.e. has a proof of IsNZ c).
  • poly_l n Q represents the injection of Q, a polynomial on n variables, as a polynomial on n+1 variables.
  • Finally, poly_s P Q : poly _ (S n) represents where P cannot mention the variable but Q can mention the variables up to and including , and the multiplication is not trivial as Q is non-null.
These indices enforce a canonical representation by ordering the multiplications of the variables. A similar encoding is actually used in the ring tactic of Coq.

Derive Signature NoConfusion NoConfusionHom for poly.
Derive Subterm for poly.

In addition to the usual eliminators of the inductive type generated by Coq, we automatically derive a few constructions on this poly datatype, and the mono datatype that follows, that will be used by the Equations command:
  • Its Signature: as described earlier , this is the packing of a polynomial with its two indices, a boolean and a natural number in this case.
  • Its NoConfusion property used to simplify equalities between constructors of the poly type (equation ).
  • Finally, its Subterm relation, to be used when performing well-founded recursion on poly.


Monomials represent parts of polynoms, and one can compute the coefficient constant by which each monomial is multiplied in a given polynom. Again the index of a mono gives the number of its free variables.

Inductive mono : nat Type :=
| mono_z : mono O
| mono_l : {n}, mono n mono (S n)
| mono_s : {n}, mono (S n) mono (S n).

Derive Signature NoConfusion NoConfusionHom Subterm for mono.

Our first interesting definition computes the coefficient in Z by which a monomial m is multiplied in a polynomial p.

Equations? get_coef {n} (m : mono n) {b} (p : poly b n) : Z by wf (pack m) mono_subterm :=
get_coef mono_z poly_z := 0%Z;
get_coef mono_z (poly_c z _) := z;
get_coef (mono_l m) (poly_l p) := get_coef m p;
get_coef (mono_l m) (poly_s p _) := get_coef m p;
get_coef (mono_s m) (poly_l _) := 0%Z;
get_coef (mono_s m) (poly_s p1 p2) := get_coef m p2.
Proof. all:repeat constructor. Defined.

The definition can be done using either the usual structural recursion of Coq or well-founded recursion. If we use structural recursion, the guardness check might not be able to verify the automatically generated proof that the function respects its graph, as it involves too much rewriting due to dependent pattern-matching. We could prove it using a dependent induction instead of using the raw fixpoint combinator as the recursion is on direct subterms of the monomial, but in general it could be arbitrarily complicated, so we present a version allowing deep pattern-matching and recursion. Note that this means we lose the definitional behavior of get_coef during proofs on open terms, but this can advantageously be replaced using explicit rewrite calls, providing much more control over simplification than the reduction tactics, especially in presence of recursive functions. The get_coef function still uses no axioms, so it can be used to compute as part of a reflexive tactic for example.
We want to do recursion on the (dependent) m : mono n argument, using the derived mono_subterm relation, which expects an element in the signature of mono, { n : nat & mono n }, so we use pack m to lift m into its signature type (pack is just an abbreviation for the signature_pack overloaded constant defined in ).
The rest of the definition is standard: to fetch a monomial coefficient, we simultaneously pattern-match on the monomial and polynomial. Note that many cases are impossible due to the invariants enforced in poly and mono. For example mono_z can only match polynomials built from poly_z or poly_c, etc.

Two detailed proofs

The monomial decomposition is actually a complete characterization of a polynomial: two polynomials with the same coefficients for every monomial are the same.
To show this, we need a lemma that shows that every non-null polynomial, has a monomial with non-null coefficient: this proof is done by dependent induction on the polynomial p. Note that the index of p rules out the poly_z case.

Lemma poly_nz {n} (p : poly false n) : m, IsNZ (get_coef m p).
Proof with (autorewrite with get_coef; auto).
  intros. depind p.
  destruct IHp. (mono_l x)...
  destruct IHp2. (mono_s x)...

Notation " ( x ; p ) " := (existT _ x p).

Theorem get_coef_eq {n} b1 b2
  (p1 : poly b1 n) (p2 : poly b2 n) :
  ( (m : mono n), get_coef m p1 = get_coef m p2)
  (b1 ; p1) = (b2 ; p2) :> { null : _ & poly null n}.
Proof with (simp get_coef in *; auto).

Throughout the proof, we use the simp tactic defined by which is a wrapper around autorewrite using the hint database associated to the constant get_coef: the database contains the defining equations of get_coef as rewrite rules that can be used to simplify calls to get_coef in the goal.

intros Hcoef.
induction p1 as [ | z Hz | n b p1 | n b p1 IHp q1 IHq ]
  in b2, p2, Hcoef |- *;
[dependent elimination p2 as [poly_z | poly_c z i] |
 dependent elimination p2 as [poly_z | poly_c z i] |
 dependent elimination p2 as
     [@poly_l n b' p2 | @poly_s n b' p2 q2] ..].
  all:(intros; try rename n0 into n; auto;
      try (specialize (Hcoef mono_z); simp get_coef in Hcoef; subst z;
           (elim i || elim Hz ||
            ltac:(repeat f_equal; auto)); fail)).
  - specialize (IHp1 _ p2). forward IHp1. intro m.
    specialize (Hcoef (mono_l m))... clear Hcoef.

We first do an induction on p1 and then eliminate (dependently) p2, the first two branches need to consider variable-closed p2s while the next two branches have p2 : poly _ (S n), hence the poly_l and poly_s patterns. The elided rest of the tactic solves simple subgoals.
We now focus on the case for poly_l on both sides. After some simplifications of the induction hypothesis using the Hcoef hypothesis, we get to the following goal:
  (b, b' : bool) (n : nat) (p1 : poly b n) (p2 : poly b' n)
  IHp1 : (b; p1) = (b'; p2)
  (b; poly_l p1) = (b'; poly_l p2)
The IHp1 hypothesis, as a general equality between dependent pairs can again be eliminated dependently to substitute b' by b and p2 by p1 simultaneously, using dependent elimination IHp1 as [eq_refl], leaving us with a trivial subgoal.

The next step is to give an evaluation semantics to polynomials. We program eval p v where v is a valuation in Z for all the variables in p : poly _ n.

Equations eval {n} {b} (p : poly b n) (v : Vector.t Z n) : Z :=
  eval poly_z nil := 0%Z;
  eval (poly_c z _) nil := z;
  eval (poly_l p) (cons _ xs) := eval p xs;
  eval (poly_s p1 p2) (cons y ys) :=
    (eval p1 ys + y × eval p2 (cons y ys))%Z.

It is quite clear that two equal polynomials should have the same value for any valuation. To show this, we first need to prove that evaluating a null polynomial always computes to 0, whichever valuation is used.
This is a typical case where the proof directly follows the definition of eval. Instead of redoing the same case splits and induction that the function performs, we can directly appeal to its elimination principle using the funelim tactic.

Lemma poly_z_eval {n} (p : poly true n) v : eval p v = 0%Z.
  funelim (eval p v); [ reflexivity | assumption ].

This leaves us with two goals as the true index in p implies that the poly_c and poly_s clauses do not need to be considered. We have to show 0 = 0 for the case p = poly_z and eval q v = 0 for the poly_l recursive constructor, in which case the conclusion directly follows from the induction hypothesis correspondinng to the recursive call. The second subgoal is hence discharged with an assumption call.
Addition is defined on two polynomials with the same number of variables and returns a (possibly null) polynomial with the same number of variables. We define an injection function to constructs objects in the dependent pair type {b : bool & poly b n}.

Definition apoly {n b} := existT (fun bpoly b n) b.

The definition shows the with feature of Equations, allowing to add a nested pattern-matching while defining the function, here in one case to inject an integer into a polynomial and in the poly_s, poly_s case to inspect a recursive call.

Notation " x .1 " := (projT1 x) (at level 3, format "x .1").
Notation " x .2 " := (projT2 x) (at level 3, format "x .2").

Equations plus {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 n) : { b : bool & poly b n } :=
  plus poly_z poly_z := apoly poly_z;
  plus poly_z (poly_c y ny) := apoly (poly_c y ny);
  plus (poly_c x nx) poly_z := apoly (poly_c x nx);
  plus (poly_c x nx) (poly_c y ny) with (x + y)%Z ⇒ {
                 | Z0apoly poly_z ;
                 | Zpos z'apoly (poly_c (Zpos z') I) ;
                 | Zneg z'apoly (poly_c (Zneg z') I) };
  plus (poly_l p1) (poly_l p2) := apoly (poly_l (plus p1 p2).2);
  plus (poly_l p1) (poly_s p2 q2) := apoly (poly_s (plus p1 p2).2 q2);
  plus (poly_s p1 q1) (poly_l p2) := apoly (poly_s (plus p1 p2).2 q1);

  plus (poly_s p1 q1) (poly_s p2 q2) with plus q1 q2 ⇒ {
       | (false ; q3)apoly (poly_s (plus p1 p2).2 q3);
       | (true ; _)apoly (poly_l (plus p1 p2).2) }.

The functional elimination principle can be derived all the same for plus, allowing us to make quick work of the proof that it is a morphism for evaluation:

Lemma plus_eval : {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 n) v,
    (eval p1 v + eval p2 v)%Z = eval (plus p1 p2).2 v.
Proof with (simp plus eval; auto with zarith).
  Ltac X := (simp plus eval; auto with zarith).
    intros until p2.
    let f := constr:(fun_elim (f:=@plus)) in apply f; intros; depelim v; X; try rewrite <- H; X.
  - rewrite Heq in Hind.
    specialize (Hind (Vector.cons h v)).
    rewrite poly_z_eval in Hind. nia.
  - rewrite Heq in Hind. rewrite <- Hind. nia.
Hint Rewrite <- @plus_eval : eval.

We skip the rest of the operations definition, poly_mult, poly_neg and poly_substract.

Equations poly_neg {n} {b} (p : poly b n) : poly b n :=
  poly_neg poly_z := poly_z;
  poly_neg (poly_c (Z.pos a) p) := poly_c (Z.neg a) p;
  poly_neg (poly_c (Z.neg a) p) := poly_c (Z.pos a) p;
  poly_neg (poly_l p) := poly_l (poly_neg p);
  poly_neg (poly_s p q) := poly_s (poly_neg p) (poly_neg q).

Lemma neg_eval : {n} {b1} (p1 : poly b1 n) v,
    (- eval p1 v)%Z = eval (poly_neg p1) v.
  Ltac XX := (autorewrite with poly_neg plus eval; auto with zarith).
  depind p1; depelim v; XX. destruct z; depelim i; XX.
  rewrite <- IHp1_1; rewrite <- IHp1_2; nia.
Hint Rewrite <- @neg_eval : eval.

Equality can be decided using the difference of polynoms
Lemma poly_diff_z_eq : {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 n),
    (plus p1 (poly_neg p2)).1 = true
    (_ ; p1) = (_; p2) :> { null : bool & poly null n }.
  depind p1; depelim p2; auto;
    try (autorewrite with poly_neg plus in H; discriminate; fail).
  - destruct z; destruct i; autorewrite with poly_neg plus in *; discriminate.
  - f_equal; destruct z as [ | z | z], z0 as [ | z0 | z0 ]; depelim i; depelim i0; autorewrite with poly_neg plus in H.
    assert (z = z0).
    remember (Z.pos z + Z.neg z0)%Z as z1; destruct z1; try discriminate; simpl in H; nia.
    subst; auto.
    remember (Z.pos z + Z.pos z0)%Z as z1; destruct z1; try discriminate.
    remember (Z.neg z + Z.neg z0)%Z as z1; destruct z1; try discriminate.
    assert (z = z0).
    remember (Z.neg z + Z.pos z0)%Z as z1; destruct z1; try discriminate; simpl in H; nia.
    subst; auto.
  - autorewrite with poly_neg plus in H.
    specialize (IHp1 _ p2 H).
    depelim IHp1. auto.
  - autorewrite with poly_neg plus in H.
    specialize (IHp1_1 _ p2_1); specialize (IHp1_2 _ p2_2).
    remember (plus p1_2 (poly_neg p2_2)) as P; remember (plus p1_1 (poly_neg p2_1)) as Q.
    destruct P as [bP P]; destruct Q as [bQ Q].
    destruct bP; destruct bQ; simpl in H; try rewrite <- HeqQ in H; try discriminate.
    specialize (IHp1_1 eq_refl); specialize (IHp1_2 eq_refl).
    depelim IHp1_1; try depelim IHp1_2; auto.

Two polynomials with the same values are syntacically equal.

This is shown using poly_nz_eval: the difference of two polynomials with the same values is null. Then use poly_diff_z_eq
Theorem poly_eval_eq : {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 n),
    ( v, eval p1 v = eval p2 v)
    (b1 ; p1) = (b2; p2) :> { b : bool & poly b n}.
  remember (plus p1 (poly_neg p2)) as P; destruct P as [b P]; destruct b.
  - apply poly_diff_z_eq; inversion HeqP; auto.
  - exfalso.
    destruct (@poly_nz_eval n) as [H0 _]; destruct (H0 P) as [v H1].
    assert (eval P v = eval (plus p1 (poly_neg p2)).2 v); [inversion HeqP; auto|].
    rewrite H2 in H1; autorewrite with eval in H1; rewrite (H v) in H1.
    rewrite IsNZ_spec in H1.

Multiplication of polynomials

This definition is a bit more laborious as there are inductive cases to treat on the second argument: it is not a simple structurally recursive definition.
The poly_l_or_s definition is a smart constructor to construct p + X × q when q can be null.

Equations poly_l_or_s {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 (S n)) :
  {b : bool & poly b (S n)} :=
poly_l_or_s p1 (b2 := true) p2 := apoly (poly_l p1);
poly_l_or_s p1 (b2 := false) p2 := apoly (poly_s p1 p2).

Lemma poly_l_or_s_eval : {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 (S n)) h v,
    eval (poly_l_or_s p1 p2).2 (Vector.cons h v) =
    (eval p1 v + h × eval p2 (Vector.cons h v))%Z.
  funelim (poly_l_or_s p1 p2); simp eval. rewrite poly_z_eval. nia.
Hint Rewrite @poly_l_or_s_eval : eval.

(* mult (poly_l p) q = mult_l q (mult p) *)

Equations mult_l {n} {b2} (p2 : poly b2 (S n)) (m : {b2} (p2 : poly b2 n), { b : bool & poly b n }) :
  { b : bool & poly b (S n) } :=
  mult_l (poly_l p2) m := apoly (poly_l (m _ p2).2);
  mult_l (poly_s p1 p2) m := poly_l_or_s (m _ p1).2 (mult_l p2 m).2.

(* mult (poly_s p1 p2) q = mult_s q (mult p1) (mult p2) *)

Equations mult_s {n} {b2} (p2 : poly b2 (S n))
     (m1 : {b2} (p2 : poly b2 n), { b : bool & poly b n })
     (m2 : {b2} (p2 : poly b2 (S n)), { b : bool & poly b (S n) }) :
    { b : bool & poly b (S n) } :=
  mult_s (poly_l p1) m1 m2 := poly_l_or_s (m1 _ p1).2 (m2 _ (poly_l p1)).2;
  mult_s (poly_s p2 q2) m1 m2 :=
    poly_l_or_s (m1 _ p2).2
                (plus (m2 _ (poly_l p2)).2 (mult_s q2 m1 m2).2).2.

Finally, the multiplication definition. This relies on the guard condition being able to unfold the definitions of mult_l and mult_s to see that multiplication is well-guarded.

Equations mult n b1 (p1 : poly b1 n) b2 (p2 : poly b2 n) : { b : bool & poly b n } :=
    mult ?(0) ?(true) poly_z b2 _ := apoly poly_z;
    mult ?(0) ?(false) (poly_c x nx) ?(true) poly_z := apoly poly_z;
    mult ?(0) ?(false) (poly_c x nx) ?(false) (poly_c y ny) :=
    match (x × y)%Z with
      | Z0apoly poly_z
      | Zpos z'apoly (poly_c (Zpos z') I)
      | Zneg z'apoly (poly_c (Zneg z') I)
    mult ?(S n) ?(b) (@poly_l n b p1) b2 q := mult_l q (mult _ _ p1);
    mult ?(S n) ?(false) (@poly_s n b p1 q1) b2 q := mult_s q (mult _ _ p1) (mult _ _ q1).
Arguments mult {n} {b1} p1 {b2} p2.

The proof that multiplication is a morphism for evaluation works as usual by induction, using previously proved lemma to get equations in Z that the nia tactic can handle.

Lemma mult_eval : {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 n) v,
    (eval p1 v × eval p2 v)%Z = eval (mult p1 p2).2 v.
Proof with (autorewrite with mult mult_l mult_s eval; auto with zarith).
  Ltac Y := (autorewrite with mult mult_l mult_s eval; auto with zarith).
  depind p1; try (depind p2; intros; depelim v; Y; simpl; Y; fail).
  depind p2; intros; depelim v; Y; simpl; Y; destruct (z × z0)%Z; simpl...
  - assert (mult_l_eval : {b2} (q : poly b2 (S n)) v h,
               eval (mult_l q (@mult _ _ p1)).2 (Vector.cons h v) =
               (eval q (Vector.cons h v) × eval p1 v)%Z).
    + depind q; intros; Y;
        rewrite <- IHp1...
      rewrite IHq2; auto; nia.
    + intros; depelim v; Y; simpl; Y; rewrite mult_l_eval...
  - assert (mult_s_eval :
               {b2} (q : poly b2 (S n)) v h,
                let mp := mult_s q (@mult _ _ p1_1) (@mult _ _ p1_2) in
                  eval mp.2 (Vector.cons h v) =
                  (eval q (Vector.cons h v) × (eval p1_1 v + h × eval p1_2 (Vector.cons h v)))%Z).
    + depind q; intros; Y; simpl; Y.
      rewrite <- IHp1_1, <- IHp1_2; Y; nia.
      rewrite <- IHp1_1. rewrite IHq2, <- IHp1_2; auto; Y; nia.
    + intros; depelim v; Y; simpl; Y; rewrite mult_s_eval...
Hint Rewrite <- @mult_eval : eval.

Boolean formulas

Armed with these definitions, we can define a reflexive tactic that solves boolean tautologies using a translation into polynomials on Z. We start with the syntax of our formulas, including variables of some type A, constants, conjunction disjunction and negation:
The have a straightforward evaluation semantics to booleans, assuming an interpretation of the variables into booleans.

Equations eval_formula {A} (v : A bool) (f : @formula A) : bool :=
  eval_formula f (f_var v) := f v;
  eval_formula f (f_const b) := b;
  eval_formula f (f_and a b) := andb (eval_formula f a) (eval_formula f b);
  eval_formula f (f_or a b) := orb (eval_formula f a) (eval_formula f b);
  eval_formula f (f_not v) := negb (eval_formula f v).

close_formula allows to obtain a formula with a fixed finite number of free variables from a formula with with variables in nat.
Definition close_formula : @formula nat { n : nat & m, m n @formula (Fin.t m) }.
  intro f; depind f.
  - unshelve eapply (S a ; _); intros m p; apply f_var.
    apply @Fin.of_nat_lt with (p := a). omega.
  - exact (O ; (fun _ _f_const b)).
  - destruct IHf1 as [n1 e1]; destruct IHf2 as [n2 e2].
    apply (existT _ (max n1 n2)); intros m p; apply f_and; [apply e1|apply e2]; nia.
  - destruct IHf1 as [n1 e1]; destruct IHf2 as [n2 e2].
    apply (existT _ (max n1 n2)); intros m p; apply f_or; [apply e1|apply e2]; nia.
  - destruct IHf as [n e].
    apply (existT _ n); intros m p; apply f_not; apply e; nia.

Definition close_formulas (f1 f2 : @formula nat) :
  { n : nat & (@formula (Fin.t n) × @formula (Fin.t n))%type }.
  destruct (close_formula f1) as [n1 e1]; destruct (close_formula f2) as [n2 e2].
  apply (existT _ (max n1 n2)); apply pair; [apply e1|apply e2]; nia.

Definitions of constant 0 poly_zero and 1 poly_one polynomials along with variable polynomials poly_var and corresponding evaluation lemmas

Fixpoint poly_zero {n} : poly true n :=
  match n with
  | Opoly_z
  | S mpoly_l poly_zero
Lemma zero_eval : n v, 0%Z = eval (@poly_zero n) v.
Proof. intros; rewrite poly_z_eval; auto. Qed.
Hint Rewrite <- @zero_eval : eval.

Fixpoint poly_one {n} : poly false n :=
  match n with
  | Opoly_c 1%Z I
  | S mpoly_l poly_one
Lemma one_eval : n v, 1%Z = eval (@poly_one n) v.
Proof. depind n; depelim v; intros; simpl; autorewrite with eval; auto. Qed.
Hint Rewrite <- @one_eval : eval.

We define an injection of variables represented as indices in Fin.t n into non-null polynoms of n variables:

Equations poly_var {n} (f : Fin.t n) : poly false n :=
  poly_var Fin.F1 := poly_s poly_zero poly_one;
  poly_var (Fin.FS f) := poly_l (poly_var f).

We can show that evaluation of the corresponding polynomial corresponds to simply fetching the value at the index in the valuation.

Lemma var_eval : n f v, Vector.nth v f = eval (@poly_var n f) v.
Proof with autorewrite with poly_var eval in *; simpl; auto with zarith.
  induction f; depelim v; intros...
Hint Rewrite <- @var_eval : eval.

Finally, we explain our interpretation of formulas as polynomials:

Equations poly_of_formula {n} (f : @formula (Fin.t n)) : { b : bool & poly b n } :=
  poly_of_formula (f_var v) := apoly (poly_var v);
  poly_of_formula (f_const false) := apoly poly_zero;
  poly_of_formula (f_const true) := apoly poly_one;
  poly_of_formula (f_not a) := plus poly_one (poly_neg (poly_of_formula a).2);
  poly_of_formula (f_and a b) := mult (poly_of_formula a).2 (poly_of_formula b).2;
  poly_of_formula (f_or a b) := plus (poly_of_formula a).2
                                          (plus (poly_of_formula b).2
          (poly_neg (mult (poly_of_formula a).2 (poly_of_formula b).2).2)).2.

The central theorem is that evaluating the formula in some valuation is the same as evaluating the translated polynomial.

Theorem poly_of_formula_eval :
   {n} (f : @formula (Fin.t n)) (v : Vector.t bool n),
    (if eval_formula (Vector.nth v) f then 1%Z else 0%Z) =
    eval (poly_of_formula f).2 (Vector.map (fun x : boolif x then 1%Z else 0%Z) v).

From this, we can derive that two boolean formulas are equivalent if the translated polynomials are themselves syntactically equal, thanks to their canonical representation.

Lemma correctness_heyting : {n} (f1 f2 : @formula (Fin.t n)),
    poly_of_formula f1 = poly_of_formula f2
     v, eval_formula (Vector.nth v) f1 = eval_formula (Vector.nth v) f2.
  intros n f1 f2 H v.
  assert (H1 := poly_of_formula_eval f1 v); assert (H2 := poly_of_formula_eval f2 v).
  remember (eval_formula (Vector.nth v) f1) as b1; remember (eval_formula (Vector.nth v) f2) as b2.
  rewrite H in H1; rewrite <- H1 in H2.
  destruct b1; destruct b2; simpl in *; (discriminate || auto).


For which theory do we have completeness? If you were attentive you might have guessed that the encodings of disjunction and conjunction are only complete for heyting boolean algebras but not classical boolean algebra, where negation is involutive.
One can avoid this problem by doing a reduction transformation on polynomials. The interested reader can look at the development for that part. Completeness can be derived for the reducing version of the translation.

Equations reduce_aux {n} {b1} (p1 : poly b1 n) {b2} (p2 : poly b2 (S n)) : { b : bool & poly b (S n) } :=
reduce_aux p1 (poly_l p2) := poly_l_or_s p1 (poly_l p2);
reduce_aux p1 (poly_s p2_1 p2_2) := poly_l_or_s p1 (plus (poly_l p2_1) p2_2).2.

Equations reduce {n} {b} (p : poly b n) : { b : bool & poly b n } :=
  reduce poly_z := apoly poly_z;
  reduce (poly_c x y) := apoly (poly_c x y);
  reduce (poly_l p) := apoly (poly_l (reduce p).2);
  reduce (poly_s p q) := reduce_aux (reduce p).2 (reduce q).2.

Theorem reduce_eval :
   {n} {b} (p : poly b n) (v : Vector.t bool n),
    eval p (Vector.map (fun x : boolif x then 1%Z else 0%Z) v) =
    eval (reduce p).2 (Vector.map (fun x : boolif x then 1%Z else 0%Z) v).
  Ltac YY := autorewrite with reduce reduce_aux eval; auto.
  depind p; intros; depelim v; YY.
  - rewrite IHp1, (IHp2 (Vector.cons h v)).
    remember (reduce p2) as P.
    destruct P as [bP P]. simpl. depelim P; simpl; YY.
    destruct h; nia.

Inductive is_reduced : {b} {n}, poly b n Prop :=
| is_reduced_z : is_reduced poly_z
| is_reduced_c : {z} {i}, is_reduced (poly_c z i)
| is_reduced_l : {b} {n} (p : poly b n), is_reduced p is_reduced (poly_l p)
| is_reduced_s : {b1} {n} (p : poly b1 n) (q : poly false n),
    is_reduced p is_reduced q is_reduced (poly_s p (poly_l q))
Derive Signature for is_reduced.

Lemma is_reduced_compat_plus : {n} {b1} (p1 : poly b1 n) (Hp1 : is_reduced p1)
                                      {b2} (p2 : poly b2 n) (Hp2 : is_reduced p2),
    is_reduced (plus p1 p2).2.
  depind Hp1; depelim Hp2; autorewrite with plus; unfold apoly; try constructor; auto.
  remember (z+z0)%Z as Z; destruct Z; constructor.
  specialize (IHHp1_2 _ q0 Hp2_2).
  remember (plus q q0) as Q; destruct Q as [bQ Q].
  destruct bQ; simpl. constructor; auto. constructor; auto.

Lemma is_reduced_compat_neg : {n} {b1} (p1 : poly b1 n) (Hp1 : is_reduced p1),
    is_reduced (poly_neg p1).
  intros. depind Hp1; try destruct z, i; autorewrite with poly_neg; try constructor; auto.

Lemma is_reduced_ok : {b} {n} (p : poly b n), is_reduced (reduce p).2.
  depind p; try constructor; auto.
  autorewrite with reduce reduce_aux.
  remember (reduce p2) as P2; destruct P2 as [bP2 P2]; depelim P2.
  destruct b0; simpl. constructor. auto. constructor; auto. depelim IHp2. auto.

  depelim IHp2. autorewrite with reduce_aux plus. unfold apoly. simpl.
  assert (R := is_reduced_compat_plus _ IHp2_1 _ IHp2_2).
  remember (plus p q) as P3; destruct P3 as [bP3 P3]. simpl.
  simpl in ×.
  destruct bP3; simpl; constructor; auto.

Lemma red_ok : {n} {b} (p : poly b n),
    is_reduced p
    ( v, eval p (Vector.map (fun x : boolif x then 1%Z else 0%Z) v) = 0%Z)
    b = true.
  intros n b p Hp H; depind Hp.
  - auto.
  - specialize (H Vector.nil); autorewrite with eval in H; destruct z, i; discriminate.
  - apply IHHp. intro v. specialize (H (Vector.cons false v)). autorewrite with eval in H. auto.
  - assert (b1 = true).
    + apply IHHp1. intro v. specialize (H (Vector.cons false v)). autorewrite with eval in H. simpl in H. rewrite Z.add_0_r in H. auto.
    + subst. apply IHHp2.
      intro v. specialize (H (Vector.cons true v)). simpl in H. autorewrite with eval in H. rewrite poly_z_eval in H. nia.

We have completeness for this form:

Lemma correctness_classical : {n} (f1 f2 : @formula (Fin.t n)),
    reduce (poly_of_formula f1).2 = reduce (poly_of_formula f2).2
     v, eval_formula (Vector.nth v) f1 = eval_formula (Vector.nth v) f2.
  intros n f1 f2; split.
  - intros H v.
    assert (H1 := poly_of_formula_eval f1 v); assert (H2 := poly_of_formula_eval f2 v).
    rewrite reduce_eval in H1; rewrite reduce_eval in H2.
    remember (eval_formula (Vector.nth v) f1) as b1; remember (eval_formula (Vector.nth v) f2) as b2.
    rewrite H in H1; rewrite <- H1 in H2.
    destruct b1; destruct b2; simpl in *; (discriminate || auto).
  - intros.
    assert ((plus (reduce (poly_of_formula f1).2).2
                  (poly_neg (reduce (poly_of_formula f2).2).2)).1 = true).
    + apply red_ok with (p := (plus (reduce (poly_of_formula f1).2).2
                                    (poly_neg (reduce (poly_of_formula f2).2).2)).2).
      × auto using is_reduced_compat_plus, is_reduced_ok, is_reduced_compat_neg.
      × intro; autorewrite with eval.
        assert (H1 := poly_of_formula_eval f1 v); assert (H2 := poly_of_formula_eval f2 v).
        rewrite <- !reduce_eval, <- H1, <- H2, (H v); nia.
    + apply poly_diff_z_eq in H0.
      remember (reduce (poly_of_formula f1).2) as P1; destruct P1 as [bP1 P1].
      remember (reduce (poly_of_formula f2).2) as P2; destruct P2 as [bP2 P2].
      destruct bP1; destruct bP2; auto; simpl in H0; depelim H0; auto.

One can check that all definitions here are axiom free, and only the proofs which depend on unfolding lemmas use the functional_extensionality_dep axiom.

Reflexive tactic

From this it is possible to derive a tactic for checking equivalence of boolean formulas. We skip the standard reification machinery and check on a few examples that indeed our tactic computes.

Ltac list_add a l :=
    let rec aux a l n :=
        match l with
          | nilconstr:((n, cons a l))
          | cons a _constr:((n, l))
          | cons ?x ?l
            match aux a l (S n) with (?n, ?l)constr:((n, cons x l)) end
        end in
    aux a l 0.

Ltac vector_of_list l :=
  match l with
  | nilconstr:(Vector.nil)
  | cons ?x ?xsconstr:(Vector.cons x xs)

Reify boolean formulas with variables in nat

Ltac read_formula f l :=
  match f with
  | trueconstr:((@f_const nat true, l))
  | falseconstr:((@f_const nat false, l))
  | orb ?x ?ymatch read_formula x l with (?x', ?l')
                    match read_formula y l' with (?y', ?l'')constr:((f_or x' y', l''))
                    end end
  | andb ?x ?ymatch read_formula x l with (?x', ?l')
                     match read_formula y l' with (?y', ?l'')constr:((f_and x' y', l''))
                     end end
  | negb ?xmatch read_formula x l with (?x', ?l')constr:((f_not x', l')) end
  | _match list_add f l with (?n, ?l')constr:((f_var n, l')) end

Ltac read_formulas x y :=
  match read_formula x (@nil bool) with (?x', ?l)
  match read_formula y l with (?y', ?l')constr:(((x', y'), l'))
  end end.

The final reflexive tactic, taking either of the correctness lemmas as argument.

Ltac bool_tauto_with f :=
  match goal with
  | [ |- ?x = ?y ] ⇒
    match read_formulas x y with
    | ((?x', ?y'), ?l)
      let ln := fresh "l" in
      let xyn := fresh "xy" in
      let nn := fresh "n" in
      let xn := fresh "x" in
      let yn := fresh "y" in
      match vector_of_list l with ?lvpose (ln := lv) end;
      pose (xyn := close_formulas x' y');
      pose (n := xyn.1); pose (xn := fst xyn.2); pose (yn := snd xyn.2);
      cbv in xyn, n, xn, yn;
      assert (H : eval_formula (Vector.nth ln) xn = eval_formula (Vector.nth ln) yn);
      [ apply f; vm_compute; reflexivity
      | exact H


Goal a b, andb a b = andb b a.
  bool_tauto_with @correctness_heyting.
Goal a b, andb (negb a) (negb b) = negb (orb a b).
  bool_tauto_with @correctness_heyting.
Goal a b, orb (negb a) (negb b) = negb (andb a b).
  bool_tauto_with @correctness_heyting.

Example neg_involutive: a, orb (negb a) a = true.
Fail bool_tauto_with @correctness_heyting.
bool_tauto_with @correctness_classical.

This page has been generated by coqdoc