At Mon, 27 Feb 2017 18:43:28 -0800, Matthew Butterick wrote: > > On Feb 27, 2017, at 5:34 PM, gfb <g...@cs.toronto.edu> wrote: > > > > If you want that to be equivalent to having `(require > > math/number-theory)` at the top-level then I'm not sure the precise > > semantics: should it affect the meaning of already-transformed code > > from before that `lifted-require` was encountered? > > Currently, I'm doing the job by hand: that is, in the expansion of the > `#%module-begin` of my #lang, I search the inbound parse tree and move the > `require` forms to the top level. This works fine. > > So I suppose my question boils down to "is there something in the > syntax-function zoo that does this more neatly?" > > As it stands, I think the answer is "no" because my manual technique is > essentially pre-empting the macro expander.
Right. The `syntax-local-lift-require` function works the way it does to avoid the potential ambiguity Gary suggests, where adding a `require` from a nested expression could create a binding that would affect expansion of surrounding forms (but it's too late for that). By adding the `require` in a fresh scope and then indirectly giving you the scope, the effect of the `require` is consistently limited to expressions that are afterwards given the new scope. At Mon, 27 Feb 2017 17:34:53 -0800 (PST), gfb wrote: > Recall that `requires` are noted before expansion of other forms, > so, e.g., this works: > > #lang racket > divides? ; Even though the require for this appears after > (require math/number-theory) Well, ok... there is some ambiguity here. It's not that `require`s are noted before other expansion, but that module-level expansions are partial until all the `defines` and `requires` are found. So `divides?` is determined not to be a definition by partial expansion, and further expansion is delayed under the assumption that `divides?` is an expression form. If `divides?` turns out to be a definition form, it will trigger an error. In contrast, if `divides?` is put after `(require math/number-theory)`, then it's ok for `divides?` to expand as a definition. This partial-expansion idea has worked ok, and so we've put up with this much order-sensitive behavior, but we try to minimize it. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
