Make Pattern_env.t private

[garrigue]

Follow-up to #12331, making Pattern_env.t private, so that we can track better who modifies it.

[Octachron]

Making Pattern_env.t private sounds like a good compromise to me.

[t6s]

We can eliminate set_env here:

let solve_constructor_annotation
    tps (penv : Pattern_env.t) name_list sty ty_args ty_ex =
  let expansion_scope = penv.equations_scope in
  let ids, new_env =
    List.fold_left
      (fun (ids, env) name ->
        let decl = new_local_type ~loc:name.loc () in
        let (id, new_env) =
          Env.enter_type ~scope:expansion_scope name.txt decl env in
        {name with txt = id} :: ids, new_env)
      ([], !!penv) name_list
  in
  let cty, ty, force =
    with_local_level ~post:(fun (_,ty,_) -> generalize_structure ty)
      (fun () -> Typetexp.transl_simple_type_delayed new_env sty)
  in
  tps.tps_pattern_force <- force :: tps.tps_pattern_force;
  let ty_args =
    let ty1 = instance ty and ty2 = instance ty in
    match ty_args with
      [] -> assert false
    | [ty_arg] ->
        unify_pat_types cty.ctyp_loc new_env ty1 ty_arg;
        [ty2]
    | _ ->
        unify_pat_types cty.ctyp_loc new_env ty1 (newty (Ttuple ty_args));
        match get_desc (expand_head new_env ty2) with
          Ttuple tyl -> tyl
        | _ -> assert false
  in
  if ids <> [] then ignore begin
    let ids = List.map (fun x -> x.txt) ids in
    let rem =
      List.fold_left
        (fun rem tv ->
          match get_desc tv with
            Tconstr(Path.Pident id, [], _) when List.mem id rem ->
              list_remove id rem
          | _ ->
              raise (Error (cty.ctyp_loc, new_env,
                            Unbound_existential (ids, ty))))
        ids ty_ex
    in
    if rem <> [] then
      raise (Error (cty.ctyp_loc, new_env,
                    Unbound_existential (ids, ty)))
  end;
  ty_args, Some (ids, cty)

[garrigue]

Added types are required in ty_args and cty, so ultimately we would still need to call Pattern_env.set_env penv new_env before returning.

[t6s]

Right. The above code indeed made two tests fail.

[garrigue]

Making the code more functional might be a good idea eventually, but this seems orthogonal to the change in Pattern_env.