[racket] How ought I run the annotate function in gui-debugger/annotator on a datum?

The documentation for `expand`:

  http://docs.racket-lang.org/reference/Expanding_Top-Level_Forms.html#%28def._%28%28quote._~23~25kernel%29._expand%29%29

says:

"Expands all non-primitive syntax in top-level-form, and returns a
syntax object for the expanded form that contains only core forms,
matching the grammar specified by Fully Expanded Programs."

The documentation for Fully Expanded Programs:

  http://docs.racket-lang.org/reference/syntax-model.html#%28part._fully-expanded%29

shows only #%plain-app -- _not_ #%app.

But when I try the example for `expand`:

    (parameterize ([current-namespace (make-base-namespace)])
      (syntax->datum
       (expand
        (datum->syntax
         #f
         '(module foo scheme
           (define a 3)
           (+ a 4))))))

I get:

    '(module foo scheme
       (#%module-begin
        (define-values (a) '3)
        (#%app call-with-values (lambda () (#%app + a '4)) print-values)))

Notice that it uses #%app -- not #%plain-app. Apparently `expand`
doesn't actually emit something that fits the spec for "Fully Expanded
Programs".

Guessing: Maybe the documentation -- and implementation -- of `expand`
predates the addition to Racket of keyword arguments?  Probably there
is some extra expansion step, which creates the "lifted" #%plain-app
wrappers for keyword argument functions I recall seeing in DrRacket's
macro stepper?

But ... how does one access this "extra" expansion step??

On Tue, Jan 21, 2014 at 9:09 PM, Patrick Boe <patrickboe at gmail.com> wrote:
> Thanks, John.
>
> As you suggest, I'm trying to do exactly what the debugger is doing.  As far
> as I can tell by tracing through its code, the debugger does the following
> to go from a raw datum ('(+ 1 2) in my example) to the syntax that it feeds
> into annotate (this is boiled down from several hundred lines of unrelated
> stuff in debug-tool.rkt)
>
>  (expand-syntax
>    (expand-syntax-to-top-form
>      (namespace-syntax-introduce
>        (datum->syntax #f
>                       '(+ 1 2)))))
>
> When I run this with racket from my command line, it fails with
>
> +: unbound identifier;
>  also, no #%app syntax transformer is bound
>   at: +
>   in: (+ 1 2)
>   context...:
>    /home/patrick/dev/scry/scree.rkt: [running body]
>
> I can wrap it with a call to (parameterize ([current-namespace
> (make-base-namespace)]) ...), but then the resulting syntax is:
>
> #<syntax (#%app + (quote 1) (quote 2))>
>
> Which, as you can see, contains #%app instead of the #%plain-app that the
> annotator is looking for.
>
> Huh.