Library prosa.util.tactics

From mathcomp Require Import ssreflect ssrbool ssrnat eqtype bigop.

Lemmas & tactics adopted (with permission) from V. Vafeiadis' Vbase.v.

Lemma neqP: (T: eqType) (x y: T), reflect (x y) (x != y).

Ltac ins := simpl in *; try done; intros.

(* ************************************************************************** *)

Exploiting a hypothesis

(* ************************************************************************** *)

Exploit an assumption (adapted from CompCert).

Lemma modusponens : (P Q : Prop), P (P Q) Q.

Ltac exploit x :=
  refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _ _) _)
  || refine (modusponens _ _ (x _ _ _) _)
  || refine (modusponens _ _ (x _ _) _)
  || refine (modusponens _ _ (x _) _).

If a subexpression expr of a goal is known to be equal to false, it may be tempting to use a lemma of the form expr = false to simply rewrite the expr with false. However, the coding style of Prosa dictates that lemmas should be stated in the form ~~ expr. Therefore, direct rewriting with such lemmas is not possible. This tactic implicitly transforms a lemma of the form ~~ expr into expr = false.
As an example, suppose we have a goal f(B1 || B2 && B3 ) = 1, where B1, B2, B3 : bool and f : bool nat. Suppose we also have several lemmas of the form Li : H1 H2 ... Hn ~~ Bi. One possible way to reduce f(B1 || B2 && B3) to f(false) is to somehow replace Bi-s with false (e.g., via negbTE) and then apply Li-s. This can be tedious if boolean variable names are long. The rewrite_neg tactic allows one to simply write rewrite_neg Bi to replace Bi with false.
Ltac rewrite_neg H :=
  let NEWH := fresh in
  (unshelve ((exploit H; last (rewrite -eqbF_neg ⇒ /eqP NEWH; rewrite NEWH; clear NEWH)) ⇒ //)) ⇒ //.

(* This tactic feeds the precondition of an implication in order to derive the conclusion
   (taken from http://comments.gmane.org/gmane.science.mathematics.logic.coq.club/7013).

   Usage: feed H.

   H: P -> Q  ==becomes==>  H: P
                            ____
                            Q

   After completing this proof, Q becomes a hypothesis in the context. *)

Ltac feed H :=
  match type of H with
  | ?foo _
    let FOO := fresh in
    assert foo as FOO; [|specialize (H FOO); clear FOO]
  end.

(* Generalization of feed for multiple hypotheses.
   feed_n is useful for accessing conclusions of long implications.

   Usage: feed_n 3 H.
     H: P1 -> P2 -> P3 -> Q.

   We'll be asked to prove P1, P2 and P3, so that Q can be inferred. *)

Ltac feed_n n H := match constr:(n) with
  | Oidtac
  | (S ?m) ⇒ feed H ; [| feed_n m H]
  end.

We introduce tactics rt_auto and fail as a shorthand for (e)auto with basic_rt_facts to facilitate automation. Here, we use scope basic_rt_facts that contains a collection of basic real-time theory lemmas. Note: constant 4 was chosen because most of the basic rt facts have the structure A1 A2 ... B, where Ai is a hypothesis usually present in the context, which gives the depth of the search which is equal to two. Two additional levels of depth (4) was added to support rare exceptions to this rule. In particular, depth 4 is needed for automatic periodic->RBF arrival model conversion. At the same time, the constant should not be too large to avoid slowdowns in case of an unsuccessful application of automation.
#[deprecated(since="prosa-0.6", note="use by or // instead")]
Ltac rt_auto := auto 4 with basic_rt_facts.
#[deprecated(since="prosa-0.6", note="use by or // instead")]
Ltac rt_eauto := eauto 4 with basic_rt_facts.

Ltac done := solve [ ssreflect.done | eauto 4 with basic_rt_facts ].
#[export] Hint Resolve I : basic_rt_facts. (* ensure the database exists *)

Note: idtac is a no-op. However, it suppresses the default obligation tactic, which uses intros to introduce unnamed variables. This is a Coq technicality that a casual reader may safely ignore. It is necessary to avoid triggering one of Prosa's continuous integration checks that validates that no proof scripts depend on automatically generated names.
#[global] Obligation Tactic := idtac.

The mathematical components library turns off Coq' s support for the enforcement of structured sub-proofs. We do want structured sub-proofs in Prosa, however, so here we turn strict checking back on.
#[global] Set Bullet Behavior "Strict Subproofs".
#[global] Set Default Goal Selector "!".

(* ************************************************************************** *)

Handier movement of inequalities.

(* ************************************************************************** *)
Ltac move_neq_down H :=
  exfalso;
  (move: H; rewrite ltnNge; move ⇒ /negP H; apply: H; clear H)
  || (move: H; rewrite leqNgt; move ⇒ /negP H; apply: H; clear H).

Ltac move_neq_up H :=
  (rewrite ltnNge; apply/negP; intros H)
  || (rewrite leqNgt; apply/negP; intros H).

The following tactic converts inequality t1 t2 into a constant k such that t2 = t1 + k and substitutes all the occurrences of t2.
Ltac interval_to_duration t1 t2 k :=
  match goal with
  | [ H: (t1 t2) = true |- _ ] ⇒
    ltac:(
      assert (EX : (k: nat), t2 = t1 + k);
      [ (t2 - t1); rewrite subnKC; auto | ];
      destruct EX as [k EQ]; subst t2; clear H
    )
  | [ H: (t1 < t2) = true |- _ ] ⇒
    ltac:(
      assert (EX : (k: nat), t2 = t1 + k);
      [ (t2 - t1); rewrite subnKC; auto using ltnW | ];
      destruct EX as [k EQ]; subst t2; clear H
    )
  | [ H: is_true(t1 t2) |- _ ] ⇒
    ltac:(
      assert (EX : (k: nat), t2 = t1 + k);
      [ (t2 - t1); rewrite subnKC; auto using ltnW | ];
      destruct EX as [k EQ]; subst t2; clear H
    )
  | [ H: is_true(t1 < t2) |- _ ] ⇒
    ltac:(
      assert (EX : (k: nat), t2 = t1 + k);
      [ (t2 - t1); rewrite subnKC; auto using ltnW | ];
      destruct EX as [k EQ]; subst t2; clear H
    )
  end.