Frankly with Inf, it's probably a good thing if your code doesn't treat two infinite values as being equal
That's a rather novel approach to overflow, and one that I would discourage.
Similarly, wrt to underflow, one could then say:
Frankly with 0, it's probably a good thing if your code doesn't treat two zero values as being equal.
my $x = 1e-5000; # 0
my $y = 1e-5100; # 0
print "equal\n" if $x==$y; # says "equal" even though mathematically they are not
That's also an approach that I would discourage.
Cheers,
Rob
| [reply] |
Well, the difference between 1e-5000 and 1e-5100 is pretty small. For most intents and purposes, there is no difference.
The difference between 6**6**6 and 7**7**7 is a lot bigger. By my calculations, the latter is 659669 orders of magnitude bigger than the former. (That is, to approximate 7**7**7, you'd need to add 659669 zeroes to the end of 6**6**6.)
| [reply] |
Well, the difference between 1e-5000 and 1e-5100 is pretty small. For most intents and purposes, there is no difference
Heh ... I thought this was about "mathematical equality", but suddenly it's now also about "intents and purposes".
For double precision NVs I could have chosen much larger values of (say) 1e-324 and 1e-325 ... but you're probably not going to be swayed by that, anyway ;-)
By my calculations, the latter is 659669 orders of magnitude bigger than the former
Yep, I get the same:
C:\>perl -MMath::MPFR -le "$x=Math::MPFR->new(6 ** 6); print 6 ** $x;
2.6591197721532269e36305
C:\>perl -MMath::MPFR -le "$x=Math::MPFR->new(7 ** 7); print 7 ** $x;
3.7598235267837884e695974
Cheers,
Rob | [reply]
[d/l] |