[bug-libmatheval] step function

[Giulio Bottazzi]

Hi,

I've got the suspicion I stumbled upon a possible bug... I think it should be step(nan)=nan. The actual implementation

double

math_step(double x)

{

/* 

* Calculate step function value.

*/

return (x < 0) ? 0 : 1;

}

of course gives step(nan)=1.

Alex, what's your opinion? And what about

double

math_step(double x)

{

/* 

 * Calculate step function value.

 */

return isnan(x) ? x : ( x < 0 ? 0 : 1 );

}

G.

--
Giulio Bottazzi <address@hidden>

PGP Key ID:BAB0A33F