Piers Cawley writes: : Damian Conway <[EMAIL PROTECTED]> writes: : > Of course, that's not to say that the particular C<undef> that's returned on : > failure-to-numerify mightn't have a property set that indicates the problem : > was not-a-numeric in nature. : : Having more than one 'undef' value sounds like a recipe for internals : madness. Or is the undef that gets slung around actually going to be a : reference to the 'real' undef?
NaN is merely the floating-point representation of undef when your variable is stored in a bare num. And if you declare a variable as int, there may well be no representation for undef at all! Similarly, it may be impossible to taint an int or a num, unless we can figure out a way to stuff such information into 0 bits. But I'd like an array of int or num to be compact. Larry