~literal vs ~datum

[Jordan Johnson]

Hi all,

Reading the documentation on ~datum in syntax/parse, I see:

The ~datum form is useful for recognizing identifiers symbolically, in contrast to the ~literal form, which recognizes them by binding.

That makes sense to me. But then I see:

> (syntax-parse (let ([define 'something-else]) #'(define x y))     [((~datum define) var:id e:expr) 'yes]     [_ 'no])

'yes

> (syntax-parse (let ([define 'something-else]) #'(define x y))     [((~literal define) var:id e:expr) 'yes]     [_ 'no])

'yes

I’m confused by this. It seems, based on the first sentence I quoted, that in the second example, in which the syntax being parsed has define bound to something other than the usual, the use of ~literal should make the first clause fail to match. And I’d have expected the examples to contrast ~datum and ~literal, rather than exhibit the same behavior.

Could someone please explain this to me?

Thanks,

Jordan

[Jens Axel Søgaard]

The binding information is attached a syntax object during expansion.

In your example the #'(define x y) haven't gone through expansion,

so you do not get what your expected result.

However we can change your macro a little:


#lang racket

(require (for-syntax syntax/parse))

(define-syntax (is-define? stx)

  (syntax-parse stx

    [(_is-define? id)

     (syntax-parse #'id

       [(~literal define) #''yes]

       [_                 #''no])]))

(is-define? define)

(let ([define 42])

  (is-define? define))

This will result in

'yes

'no

/Jens Axel

[Jordan Johnson]

On Aug 1, 2017, at 11:34, Jens Axel Søgaard <jens...@soegaard.net> wrote:

The binding information is attached a syntax object during expansion.

In your example the #'(define x y) haven't gone through expansion,

so you do not get what your expected result.

However we can change your macro a little...

Ah, that clarifies everything. Thanks!

So, the example I quoted is actually from the docs on ~datum, and it seems pretty clear that it’s trying unsuccessfully to elicit different behaviors that show the difference between ~literal and ~datum. Based on the code you’ve provided, I think this would be a better illustration:

(define-syntax (is-define? stx)

  (syntax-parse stx

    [(_is-define? id)

     (syntax-parse #'id

       [(~literal define) #''yes]

       [(~datum   define) #''not-really]

       [_                 #''not-even-close])]))

(is-define? define)         ;; 'yes

(let ([define 42])

  (is-define? define))      ;; 'not-really

(is-define? something-else) ;; 'not-even-close

I’ve submitted that as a PR on the doc file linked above. I hope you don’t mind my using your example in that way. It was very helpful.

Regards,

Jordan

[Shu-Hung You]

I think the case here is a little subtle. The syntax object #'(define
x y) has actually gone through expansion, but Racket's new expansion
model will prune this syntax object's local scopes [1]. One could use
#:local [2] in this case to see the difference:

> (syntax-parse (let ([define 'something-else]) (quote-syntax (define x y)))

[((~literal define) var:id e:expr) 'yes]
[_ 'no])
'yes

> (syntax-parse (let ([define 'something-else]) (quote-syntax (define x y) #:local))

[((~literal define) var:id e:expr) 'yes]
[_ 'no])

'no

The original doc predates the new expander and the result was probably
'no in the old expansion model.

Best,
Shu-Hung

1. http://www.cs.utah.edu/plt/scope-sets/general-macros.html#%28part._.Local_.Bindings_and_.Syntax_.Quoting%29

2. https://docs.racket-lang.org/reference/Syntax_Quoting__quote-syntax.html

> --
> You received this message because you are subscribed to the Google Groups
> "Racket Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to racket-users...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.