Release 1.2beta of Equations is available, as source and through opam (packages coq-equations.1.2~beta+8.8 and coq-equations.1.2~beta+8.9 for Coq 8.8 and Coq 8.9).

I am pleased to announce release 1.2beta of the Equations package. Equations is a function definition plugin for Coq, that supports the definition of functions by dependent pattern-matching and structural or well-founded, mutual or nested recursion and compiles them into core Coq terms without axioms. It automatically derives the clauses’ equations, the graph of the function and its associated elimination principle.

This version of Equations is based on an improved simplification engine for the dependent equalities appearing during dependent eliminations that is also usable as a separate dependent elimination tactic, providing an axiom-free variant of dependent destruction and a more powerful form of inversion. Along with bug fixes, this version introduces a few important changes:

  • At the syntax level, patterns on the left-hand sides of definitions are now interpreted using Coq’s standard term interpreter, so that they naturally support implicit arguments and notations. Recursive and local notations are also available. Finally one does not have to introduce binders for arguments in the function’s prototype, e.g.:

    Equations plus : nat -> nat -> nat :=
{ 0   + m := m;
  S n + m := S (n + m) }
where "x + y" := (plus x y).

  • The annotations for recursive programs are now written using a by annotation before the program’s clauses, as in:

    Equations gcd (x y : nat) : nat by wf (x + y) lt :=
gcd 0 x := x ;
gcd x 0 := x ;
gcd x y with gt_eq_gt_dec x y := {
    | inleft (left ygtx) := gcd x (y - x) ;
    | inleft (right refl) := x ;
    | inright xgty := gcd (x - y) y }.

    The reference manual defines the input syntax precisely.

  • An homogeneous no-confusion principle on indexed families can be automatically derived for those families whose no-confusion property does not require uniqueness of identity proofs. This includes, e.g. fin, vector and many other indexed data structures (examples: definterp, polynomials).

  • Support for nested well-founded recursion, which naturally express standard patterns of functional programming like the worker/wrapper pattern (example: accumulator).

  • Mutual well-founded recursive definitions can be implemented by mixing well-founded recursion and dependent pattern-matching (example: mutualwfrec).

  • Dependent where clauses help incrementally build dependently-typed values. A notation for pattern-matching lambdas, which are elaborated to where clauses is also available (example: definterp).

  • An experimental refinement mode opens the proof mode to fill the holes of an Equations program, making expected types and contexts of partial definitions easier to see, and lets one seamlessly switch between refinement and proof mode. BEWARE: abstract does not work well in that case. The syntax to switch to this mode is currently Equations?. Otherwise, the obligations are generated as for Program.

  • A relocatable library of support lemmas allows one to use Equations in the context of HoTT-Coq or to benefit from strict propositions (WIP).

The main features of Equations include:

  • Dependent pattern-matching in the style of Agda/Epigram, with inaccessible patterns, with and where clauses. The use of the K axiom or a proof of K is configurable.

  • Support for mutual and nested well-founded recursion using by wf annotations, and automatic derivation of the subterm relation for inductive families.

  • Support for mutual and nested structural recursion using with and where auxiliary definitions, allowing one to factor multiple uses of the same nested fixpoint definition. It proves the expected elimination principles for mutual and nested definitions.

  • Automatic generation of the defining equations as rewrite rules for every definition.

  • Automatic generation of the unfolding lemma for well-founded definitions (requiring the functional extensionality axiom).

  • Automatic derivation of the graph of the function and its elimination principle.

  • A new dependent elimination tactic based on the same splitting tree compilation scheme that can advantageously replace dependent destruction and inversion. The as clause of dependent elimination can be used to specify exactly the patterns and naming of new variables needed by an elimination.

  • A set of Derive commands for automatic derivation of constructions from an inductive type: its signature, no-confusion properties, well-founded subterm relation and decidable equality proof, if applicable.

The current system is usable for developing non-toy programs and proofs (see examples), although it still has some limitations. Documentation is available on the website, including a reference manual with an introduction to the features of the plugin. Feedback is very welcome!

– Matthieu Sozeau