Here's how various Schemes deal with syntax for non-finite inexact numbers. "Standard syntax" means what R6RS prescribes: +inf.0 for positive infinity, -inf.0 for negative infinity, and +nan.0 (or -nan.0) for NaN. The current draft of R7RS does not prescribe -nan.0; it is under ballot at this writing.
Racket, Gauche, Chicken (with or without the numbers egg), Bigloo, Scheme48, Guile, Kawa, Chez, Ikarus/Vicare, Larceny, Ypsilon, Mosh, IronScheme, STklos, Spark accept and print the standard syntax.
Gambit, Chibi accept and print the standard syntax, except that they do not accept -nan.0.
SigScheme, Scheme 9, Dream, Oaklisp, Owl Lisp are excluded because they do not have inexact non-finite numbers.
MIT has full control over floating-point exceptions. To ignore exceptions, one must do (flo:ignore-exception-traps thunk).
The following table concisely describes the other Schemes in the test suite. "Std syntax" is "yes" if the Scheme can read the standard syntax, "print" shows what (let* ((i (* 1.0e200 1.0e200)) (n (- i i))) (list i (- i) n)) prints, and "own syntax" is "yes" if the Scheme can reread what it prints.
|
Scheme |
std syntax |
prints |
own syntax |
|
TinyScheme |
no |
(inf -inf -nan) |
no |
|
BDC |
no |
(Infinity -Infinity NaN) |
no |
|
Elk |
no |
(inf -inf -nan.0) |
no |
|
Inlab |
no |
(inf.0 -inf.0 -nan.0) |
no |
|
KSi |
yes |
(+inf.0 -inf.0 nan.0) |
yes |
|
MIT |
no |
** |
yes |
|
NexJ |
yes |
(Infinity -Infinity NaN) |
no |
|
Rep |
no |
(inf. -inf. -nan.) |
no |
|
RScheme |
no |
(inf. -inf. -nan.) |
no |
|
Scheme 7 |
no |
(inf.0 -inf.0 -nan.0) |
no |
|
Schemik |
no |
(inf -inf -nan) |
no |
|
SCM |
* |
(+inf.0 -inf.0 0/0) |
yes |
|
scsh |
no |
(inf. -inf. -nan.) |
no |
|
Shoe |
no |
(inf -inf -nan) |
no |
|
SISC |
no |
(infinity.0 -infinity.0 nan.0) |
no |
|
SXM |
no |
(inf.0 -inf.0 -nan.0) |
no |
|
UMB |
no |
(inf.0 -inf.0 -nan) |
no |
|
VX |
yes |
(inf. -inf. -nan.) |
no |
|
XLisp |
no |
(inf -inf -nan) |
no |
[*] Accepts +inf.0 and -inf.0 but not +nan.0 or -nan.0
[**] Prints #[+inf] and #[-inf] for infinities. Will add info on NaNs when I have it. -- ArthurGleckler