Expose local phase levels to rules

[jackfirth]

This rule:

(define-refactoring-rule redundant-or
  #:description "This `or` expression does nothing and can be replaced with its subexpression."
  #:literals (or)
  (or a1:id a2:id)
  #:when (free-identifier=? #'a1 #'a2)
  a1)

Fails to refactor this code:

(begin-for-syntax
  (define x 42)
  (or x x))

The reason for that is that free-identifier=? checks if two identifiers have the same binding at a specific phase level, and by default it uses (syntax-local-phase-level). This is reasonable for macros, which set (syntax-local-phase-level) to the level of the macro invocation. But refactoring rules aren't run during macro expansion, so (syntax-local-phase-level) always returns zero. This lets free-identifier=? in rules work fine on phase 0 code, but it fails on phase 1 code such as code inside begin-for-syntax.

Resyntax ought to expose the local phase level to rules. I'm not sure how to get that information out of the expander using the current expansion observer hook, but it's probably possible. One way to expose it to rules might be to add a (resyntax-local-phase-level) parameter and let rules use that to access the phase level. On the Resyntax source analysis side, there could be an expansion-visit struct / record type containing the syntax object that the expander visited and the phase level the visit occurred at.

[jackfirth]

Hmm. Just using the local phase level makes it tricky if a rule wants to check free-identifier=? on an identifer within a (begin-for-syntax ...) form given to the rule, since the phase the identifier was resolved at during expansion is not the same phase that the rule's code was visited at. Perhaps instead of exposing a (resyntax-local-phase-level) parameter, Resyntax could attach a 'visit-phase syntax property to each syntax object (and its subforms) to track what phase it was used at.

However, the same identifier could end up being used at multiple phases if a macro duplicates it. So maybe (resyntax-local-phase-level) really is the right way to go. I'll have to think about this more.