expander: add syntax-local-make-definition-context-introducer

[usaoc]

Checklist

• Feature
• tests included
• documentation

Description of change

This function makes an introducer that encapsulates a scope that should be pruned from quote-syntax forms, much like an {inside,outside}-edge scope. This can be useful for implementing forms like Rhombus let (which is indeed the motivation; see racket/rhombus#471).

[usaoc]

To clear up potential confusion seeing the mentioning Rhombus issue: this PR does not intend to introduce low-level mechanism for the user to replicate what definition contexts already do. Rather, it’s intended for situation where you want to create scopes that should be pruned for quote-syntax, so it’s limited in scope.

[mflatt]

Since the as-inside-edge? argument affects only the debugging name of the scope, would it make sense to generalize that argument to be any symbol that isn't 'macro?

Specifically, I have in mind that this function is useful to add use-site scopes to Rhombus's enforestation-level expansion. The name doesn't matter, but it can help with debugging if things are distinct. Also, seeing a debugging category for Rhombus let by be helpful.

Changing to a symbol would also help clarify that debugging is the only effect of the choice.

[mflatt]

As It may also be that syntax-local-apply-transformer is the better way to handle use-site scopes in Rhombus, as @michaelballantyne points out in racket/rhombus#474. But if we're going to expose the more primitive building block directly, then the generalization away from as-inside-edge?.

[usaoc]

I incorporated the suggestion and made some clarification to the purpose of this function. I’m not sure how much we still want this change, though.

[mflatt]

Thanks for the updates!

Making sure I'm not confused: This is still a useful shortcut for implementing something like let in Rhombus, right? Even with Rhombus using syntax-local-apply-transformer, the scope for each let is separate from the ones managed directly by syntax-local-apply-transformer.

[usaoc]

Yes, if we keep the current implementation strategy for let, that is, managing a single scope for each let.