Every module is indeed compiled with a fresh context. The untyped
module doesn't directly refer to typed-racket/utils/tc-utils, which is
the file that defines typed-context?, but it does refer to a typed
module, whose expansion will contain a reference to that module.
Consider:
#lang typed/racket
(provide f)
(define (f [x : Integer]) (+ x 1))
That expands to (among a lot of other things):
(define-values (f) (lambda (x) (#%app + x '1)))
(define-syntaxes (f.1) (#%app make-redirect2 (quote-syntax f.3)))
(define-syntaxes
(f.2)
(#%app make-typed-renaming (t-quote-syntax f) (t-quote-syntax f.1)))
Where f.2 is provided as f, and f.3 is the contracted version of the identifier.
Then `make-typed-renaming` is effectively `(lambda (a b) (lambda (stx)
(if (unbox typed-context?) a b)))`
So it's the definition of `f` that depends on the box, and so there's
a fresh instantiation of `tc-utils` for every module that gets
expanded that depends on the typed module above.
Hopefully that clarifies things a bit more.
Sam
On Wed, Oct 7, 2020 at 10:07 PM Nathaniel Griswold
<[email protected]> wrote:
>
> Hm, I have this two reads for a over and I didn’t quite get it. I thought
> every module was compiled in a fresh context so my issue is I’m wondering how
> that identifier for the box is defined for syntax expansion in untyped
> modules. Maybe you understand what I mean; otherwise I’ll give the paper a
> look later tonight or tomorrow to see what I’m missing here.
>
> Thanks
>
> > On Oct 7, 2020, at 8:44 PM, Sam Tobin-Hochstadt <[email protected]>
> > wrote:
> >
> > The way this works, the box doesn't have to "get into" any modules
> > that aren't typed. Instead, when expanding a module A, that module
> > may contain uses of identifiers defined in a typed module T. Those
> > identifiers are macros that look something like this:
> >
> > (define-syntax (id stx) (if typed-context? #'typed-id #'id-for-untyped))
> >
> > Furthermore, the typed-context? value is initially set to #f, and it's
> > set to #true only inside the typechecker. When `id` is used in an
> > untyped module, it never gets set to #true, and so the
> > `id-for-untyped` is used. That means that the untyped code doesn't
> > need to know or do anything to make the system work, which is
> > necessary for Typed Racket to be a library.
> >
> > There's more detail about this in our Scheme Workshop paper, here:
> > https://www2.ccs.neu.edu/racket/pubs/scheme2007-ctf.pdf
> >
> > Sam
> >
> >
> >> On Wed, Oct 7, 2020 at 6:18 AM Nate Griswold <[email protected]>
> >> wrote:
> >>
> >> I just finished reading the paper "Languages as Libraries", which really
> >> helped as an introduction to racket's language system (i had previously
> >> been ignoring the syntax features).
> >>
> >> I did have one question from it, which wasn't quite made clear in the
> >> interoperability section, or so it seems: how does the box
> >> `typed-context?` get into the modules that are not part of that language?
> >> Or more generally how would you solve this problem of needing to know
> >> which language of a module you are in without modifying the other
> >> languages? I was wondering if it was a hackaround or if there was a way to
> >> do generically in any language i might make.
> >>
> >> I grepped the source code and it looks like it's even exposed to the user
> >> in a function `syntax-local-typed-context?`, but i noticed that module
> >> never actually imports `typed-context?`.
> >>
> >> I suppose if it is a hack i could just reflect on the namespace to solve
> >> the problem, but i thought i would send a mail here to check.
> >>
> >> Thanks for the great work everyone!
> >>
> >> Nate
> >>
> >> --
> >> You received this message because you are subscribed to the Google Groups
> >> "Racket Developers" group.
> >> To unsubscribe from this group and stop receiving emails from it, send an
> >> email to [email protected].
> >> To view this discussion on the web visit
> >> https://groups.google.com/d/msgid/racket-dev/CAM-xLPpey6e2S_JsBi_ZA_0OAC%2Bh8NJDrVYRv63fWmYrWiig1A%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups
"Racket Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/racket-dev/CAK%3DHD%2BavJJ5Yrov1fLEoDJO9j01QGGbqH%2BkJdFHuQMjDdj5e4Q%40mail.gmail.com.
