GitLab

Not only Theory, but also Task and Transformation 
do not need to depend on environment. Moreover,
we don't have to track the clone history in tasks.

The only thing we need to do, is to provide three
registration functions in Driver, namely:

  val register_transform : string -> (unit -> task_t) -> unit
  val register_transform_env : string -> (env -> task_t) -> unit
  val register_transform_clone : string -> (env -> clone -> task_t) -> unit

and then another three for task_list_t.

Then any particular transformation that is going to depend
on environment or clone_history, must register itself via
the appropriate registration function. It will be the
responsibility of Driver to recreate the transformation
for every new environment and/or clone history. It will
be easy, given that both [env] and [clone] are physically
comparable and provide a unique tag.

Thus, the generic interface provided by Transform can be
completely independent on [env] and [clone].

This commit implements the proposed interface of Task
and moves the environment stuff into a separate module.

Experimentation sur l'interface de transform.ml. Par contre quand est ce que le contexte est applati dans theory.ml?

Mise à jour de transform, inlining,flatten,... cependant anomalie car un use peut être ajouté par add_decl...