[racket-dev] enum returns #f on c->racket no match?

From: John Clements (clements at brinckerhoff.org)
Date: Tue Nov 16 14:02:07 EST 2010

On Nov 15, 2010, at 3:02 PM, John Clements wrote:

> The documentation for '_enum" says this:
> 3.8 Enumerations and Masks
> Although the constructors below are describes as procedures, they are implemented as syntax, so that error messages can report a type name where the syntactic context implies one.
> (_enum symbols [basetype]) → ctype?
>  symbols : list?
>  basetype : ctype? = _ufixint
> Takes a list of symbols and generates an enumeration type. The enumeration maps between a symbol in the given symbols list and corresponding integers, counting from0.
> The list symbols can also set the values of symbols by putting '= and an exact integer after the symbol. For example, the list '(x y = 10 z) maps 'x to 0, 'y to 10, and 'z to11.
> The basetype argument specifies the base type to use.
> This says nothing about what happens when using racket->C on a symbol that's not mentioned by the enumeration, or what happens when using C->racket on a number that's not mentioned in the enumeration. 
> Based on my tests, it appears that the conversion signals an error in the racket->C direction, but simply returns #f in the C->racket direction.  
> Should I document the current behavior, or would it make more sense to change it to signal an error rather than returning #f?

Okay the existing behavior was bothering me because (post-conversion to #f) there's no way to recover the problematic integer.  I therefore changed it so it signals an error instead, like this:

. . . plt/collects/racket/private/more-scheme.rkt:265:2: enum:int->_my-enum: expected a known integer from C, got: 6

That error comes from linking to a file with this C function:

int tester(int x){
  return x+2;

... using this racket program:

#lang racket

(require ffi/unsafe)

(define lib (ffi-lib "/tmp/tester.dylib"))

(define _my-enum
   '(chicken = 3
     monkey = 4
     duck = 5)))

(define tester
  (get-ffi-obj "tester" lib (_fun _my-enum -> _my-enum)))

(tester 'monkey)

... and finally, here's the diff:

pcp062767pcs:~/plt/collects/ffi clements$ git diff unsafe.rkt
diff --git a/collects/ffi/unsafe.rkt b/collects/ffi/unsafe.rkt
index 66fd34a..99e0d7f 100644
--- a/collects/ffi/unsafe.rkt
+++ b/collects/ffi/unsafe.rkt
@@ -765,6 +765,8 @@
   (define int->sym '())
   (define s->c
     (if name (string->symbol (format "enum:~a->int" name)) 'enum->int))
+  (define c->s
+    (if name (string->symbol (format "enum:int->~a" name)) 'int->enum))
   (let loop ([i 0] [symbols symbols])
     (unless (null? symbols)
       (let-values ([(i rest)
@@ -784,7 +786,11 @@
         (if a
           (cdr a)
           (raise-type-error s->c (format "~a" (or name "enum")) x))))
-    (lambda (x) (cond [(assq x int->sym) => cdr] [else #f]))))
+    (lambda (x) (cond [(assq x int->sym) => cdr] 
+                      [else
+                       (error c->s 
+                              "expected a known integer from C, got: ~s"
+                              x)]))))
 ;; Macro wrapper -- no need for a name
 (provide _enum)

Let me know if it's okay to commit this.  If I don't hear back in a couple of days, I'll just go ahead :).


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4669 bytes
Desc: not available
URL: <http://lists.racket-lang.org/dev/archive/attachments/20101116/c4529173/attachment.p7s>

Posted on the dev mailing list.