Stick the identifier into a define-syntax and retrieve it elsewhere. My idiom is
> #lang racket > > (define-for-syntax it #f) > > (begin-for-syntax > (set! it #'a)) > > (define-syntax (m stx) > (syntax-case stx () > [(_ . x) it])) > > (define a 10) > > (m) I have used this on one occasion. Does it work for you? On Jul 20, 2014, at 6:27 PM, Asumu Takikawa <as...@ccs.neu.edu> wrote: > Hi all, > > I'm stuck on a macro issue with the top-level and was wondering if > anyone had any thoughts. > > Certain macros (I'm thinking of parts of Typed Racket) manage > communication between different parts of a program (different macros or > different instantiations of the same macro) by storing a syntax-time > table that maps, say, identifiers that should be defined in one place > and referenced in another. > > The trouble is that this approach doesn't work so well at the top-level > because of the issues described in Reference ยง1.2.3.9. Namely because at > the top-level if one invocation of a macro expands to: > > (my-macro) => > > (begin > (define-values (generated-id) something-to-use-later) > ... other stuff ...) > > and then another invocation expands to: > > (my-macro) => > > (begin > (generated-id) > ... other stuff ...) > > you get an unbound identifier error because `generated-id` is only bound > within the first `begin` block. > > Is there any way I can get around this issue? I thought maybe syntax > lifting would help out, but I need to use a particular generated > identifier and not a fresh one obtained from lifting. That and TR > catches all lifts at the top-level anyway (and sticks them in a `begin`) > so I don't think it'll help. > > I also noticed that you can get around the 1.2.3.9 limitation using > `eval` if you have a non-gensymed identifier, but it doesn't appear to > help with a gensymed one. > > Cheers, > Asumu > ____________________ > Racket Users list: > http://lists.racket-lang.org/users ____________________ Racket Users list: http://lists.racket-lang.org/users
