Question about syntax-local-lift-require

Questions & Answers
1

Hi,

The documentation of syntax-local-lift-require says

If raw-require-spec and stx are part of the input to a transformer, then typically syntax-local-introduce should be applied to each before passing them to syntax-local-lift-require, and then syntax-local-introduce should be applied to the result of syntax-local-lift-require. Otherwise, marks added by the macro expander can prevent access to the new imports.

But the two applications of syntax-local-introduce to stx will just cancel each other, so what's the purpose of applying syntax-local-introduce to stx?

2

In the interval between the two points in time, the expander views potential identifiers as "syntax-locally-introduced” and acts accordingly. But that’s of course not the reality. So at the end, your macro removes these scopes again to set up the proper bindings.

3

Perhaps it was necesary when syntax marks were not conmutative??

4

What does it mean for syntax marks to be noncommutative?

5

I see your point. Except as the function description says (and I checked the function's source code to confirm), all syntax-local-lift-require does is (1) do a require lift, which does not use stx in anyway, and (2) add a scope to stx and return it. So it doesn't matter if stx is "syntax-locally-introduced” or not within syntax-local-lift-require.

6

For example in this program:

#lang racket

(define id #'hello)

(define mark1 (make-syntax-introducer))
(define mark2 (make-syntax-introducer))

(bound-identifier=? (mark1 (mark2 (mark2 (mark1 id)))) id)
(bound-identifier=? (mark1 (mark2 (mark1 (mark2 id)))) id)

In the curent version of Rackert, both results are #t.

IIRC up to version 6.2 the results were #t and #f, because marks could only remove a "consecutive" identical mark.

1 Like
7

I see. I think syntax-marks are commutative now, as the Racket Reference make it pretty clear that the scopes associated with a syntax object form a scope set.