[plt-scheme] Using syntax before define-syntax is weird

You can also use #%expression (typically when the call to bob was an
argument to the macro).

Robby

On Wed, Apr 7, 2010 at 11:39 PM, Carl Eastlund <cce at ccs.neu.edu> wrote:
> On Thu, Apr 8, 2010 at 12:16 AM, Neil Toronto <ntoronto at cs.byu.edu> wrote:
>> DrScheme 4.2.5. This:
>>
>> (define-syntax (bob stx)
>>  (syntax-case stx ()
>>    [(_)  #'3]
>>    [_  #'4]))
>>
>> (bob)
>>
>>
>> outputs 3 as it should. This:
>>
>> (bob)
>>
>> (define-syntax (bob stx)
>>  (syntax-case stx ()
>>    [(_)  #'3]
>>    [_  #'4]))
>>
>>
>> results in "procedure application: expected procedure, given: 4 (no
>> arguments)" The macro stepper verifies that (bob) is getting expanded to
>> (4). IOW, the expander never passes #'(bob) to "bob".
>>
>> Is this intended behavior?
>>
>> Neil T
>
> Yup.  There are two passes to expansion: head expansion, then full
> expansion.  The expander has to find out if "bob" defines syntax that
> the subsequent form will use, or variables it will reference.  So it
> expands it far enough to find out if it's a definition or expression,
> but doesn't do anything more with the body unless it's a syntax
> definition.
>
> In this case, it doesn't know what "bob" is during head expansion, so
> it gives up.  If you had (list (bob)), it'd work fine.  It'd say "oh,
> list?  that's an expression", save the rest of the expression
> expansion for later, and get to the define-syntax.  Then when it got
> back to the rest of the expression, it'd know all about bob.
>
> --Carl
> _________________________________________________
>  For list-related administrative tasks:
>  http://list.cs.brown.edu/mailman/listinfo/plt-scheme
>