GitLab

Still rely on String_UIP.

definition of types

Even for Assoc, which is a bit heavy but works well.

coerce is now a smart constructor.

Examples pass.

select and update

Coercion cannot be defined as a function, counterexamples added to
test.sigma.

We have to distinguish objects (ς-terms) and expressions (ς-terms with
explicit coercions).

In order to get an expression from an object, subtyping should be
reflexive, which is only the case for well-formed types (associations
lists without duplicates) so we add an invariant on types using a
Σ-type.

Because of these, reduction in Coq is very slow and Coq cannot check the
examples yet.

Needed for the PromCell example