[racket-dev] syntax-local-inferred-name, keyword functions, and syntax properties

On Mon, Dec 9, 2013 at 4:50 PM, Robby Findler
<robby at eecs.northwestern.edu> wrote:
> I'm not really familiar with the way the keyword expansion works. I was just
> trying to suggest "obvious" reasons why the code would do such a thing.
>
> But if you find that changing the name changes the error message, then
> changing that would be bad.

The point I'm trying to make is that I think the error message is
_currently bad_, because there's no reason to think of that procedure
as named `foo`. We'd be better off and less confused if the procedure
in the error message didn't have a name.

If you think changing it would be bad, why do you think the relevant
procedure (the one with a keyword named `#:kk`) should be named `foo`?

Sam

>
> Sorry for the noise.
>
> Robby
>
>
> On Mon, Dec 9, 2013 at 3:41 PM, Sam Tobin-Hochstadt <samth at cs.indiana.edu>
> wrote:
>>
>> Yes, that would work, but I still don't see why that's a useful name to
>> use.
>>
>> Sam
>>
>> On Mon, Dec 9, 2013 at 4:37 PM, Robby Findler
>> <robby at eecs.northwestern.edu> wrote:
>> > Would it work to make it use syntax-local-infer-name but only use the
>> > symbolic part of that name?
>> >
>> > Robby
>> >
>> >
>> >
>> > On Mon, Dec 9, 2013 at 3:34 PM, Sam Tobin-Hochstadt
>> > <samth at cs.indiana.edu>
>> > wrote:
>> >>
>> >> Aha, I found the example:
>> >>
>> >>    (let ([foo ((lambda (#:kk x) (λ (x) x)) #:k 0)]) (foo 5))
>> >>
>> >> Has this error:
>> >>
>> >> application: procedure does not expect an argument with given keyword
>> >>   procedure: foo
>> >>   given keyword: #:k
>> >>   arguments...:
>> >>    #:k 0
>> >>
>> >>
>> >> Note that if you make the keywords the same, `foo` is indeed bound to
>> >> a procedure, but _not_ to the procedure referred to in the error
>> >> message.
>> >>
>> >> Sam
>> >>
>> >> On Mon, Dec 9, 2013 at 4:29 PM, Sam Tobin-Hochstadt
>> >> <samth at cs.indiana.edu> wrote:
>> >> > I haven't found a way to make it happen yet.  But even so, it seems
>> >> > like the wrong name.
>> >> >
>> >> > Sam
>> >> >
>> >> > On Mon, Dec 9, 2013 at 4:16 PM, Robby Findler
>> >> > <robby at eecs.northwestern.edu> wrote:
>> >> >> Is it possible that that name can leak out in an error message?
>> >> >>
>> >> >> Robby
>> >> >>
>> >> >>
>> >> >>
>> >> >> On Mon, Dec 9, 2013 at 3:10 PM, Sam Tobin-Hochstadt
>> >> >> <samth at cs.indiana.edu>
>> >> >> wrote:
>> >> >>>
>> >> >>> Currently, this program fails with a somewhat bizarre type error:
>> >> >>>
>> >> >>>     #lang typed/racket
>> >> >>>
>> >> >>>     (: foo ([#:k Any] -> Integer))
>> >> >>>     (define (foo #:k [s #f]) 0)
>> >> >>>
>> >> >>>     (let: ([i : Integer (foo #:k #t)]) i)
>> >> >>>
>> >> >>> The reason is that the expansion of keyword applications generates
>> >> >>> a
>> >> >>> name to use for the function (here `foo`), and it uses
>> >> >>> `syntax-local-infer-name` to get the name to use. Unfortunately, in
>> >> >>> this case, it produces `i`, an identifier which has an extra syntax
>> >> >>> property saying that `i` is an `Integer`.  Of course, `foo` isn't
>> >> >>> an
>> >> >>> integer, it's a function, and so we get a type error.
>> >> >>>
>> >> >>> I don't see why the inferred name is the right choice here --
>> >> >>> there's
>> >> >>> no connection between `i` and the name of the function. I can just
>> >> >>> change this to use a fresh name, but I thought I'd ask first.
>> >> >>>
>> >> >>> Sam
>> >> >>> _________________________
>> >> >>>   Racket Developers list:
>> >> >>>   http://lists.racket-lang.org/dev
>> >> >>
>> >> >>
>> >
>> >
>
>