Re^5: Small Perl 6 discoveries II, Rats

in reply to Re^4: Small Perl 6 discoveries II, Rats
in thread [Perl6] Small discoveries I, __DATA__

> perl5 -E 'say 3e-1 - .3'
0
> perl6 -e 'say 3e-1 - .3'
5.55111512312578e-17
[download]

Comment on Re^5: Small Perl 6 discoveries II, Rats Download Code

Replies are listed 'Best First'.
Re^6: Small Perl 6 discoveries II, Rats by Laurent_R (Canon) on Sep 28, 2017 at 14:11 UTC
`> perl5 -E 'say 3e-1 - .3' 0` [download] In this case, the floating values for 3e-1 and .3 are both "wrong," i.e. not really equal to mathematical .3, but since they have the same error, subtracting one from the other happens to return 0. So the result seems correct, but only because the two inaccuracies compensate each other. `> perl6 -e 'say 3e-1 - .3' 5.55111512312578e-17` [download] Here, voluntarily or not, you mix up two different types. 3e-1 is a floating-point value (and, as such, liable to inaccuracies, as we all know), whereas .3 is a rational, and is entirely accurate. For example, printing the first 100 digits of .3 under Perl 6: `> printf "%.100f\n", .3; 0.30000000000000000000000000000000000000000000000000000000000000000000 +00000000000000000000000000000000 >` [download] Subtracting a rational from an inaccurate float is just wrong, it is like adding apples and oranges.	[reply] [d/l] [select]
Re^7: Small Perl 6 discoveries II, Rats by Anonymous Monk on Sep 28, 2017 at 14:22 UTC
Am I allowed to subtract a rational from a float if I explicitly convert it to a float first? Pretty please? `> 3e-1 - Num(.3) 5.55111512312578e-17` [download] Oh look, I got the same answer. Also, Perl6's `printf` lies. `> printf "%.100f\n", .3 0.30000000000000000000000000000000000000000000000000000000000000000000 +00000000000000000000000000000000 > printf "%.100f\n", 3e-1 0.30000000000000000000000000000000000000000000000000000000000000000000 +00000000000000000000000000000000 > printf "%.100f\n", Num(.3) 0.30000000000000000000000000000000000000000000000000000000000000000000 +00000000000000000000000000000000` [download] But you can still get an honest `printf` with `Inline::Perl5`. `use Inline::Perl5; my $p5 = Inline::Perl5.new; $p5.run('sub p { printf "%.100f\n", shift }'); $p5.call('p', .3); $p5.call('p', Num(.3)); $p5.call('p', 3e-1); =finish 0.29999999999999998889776975374843459576368331909179687500000000000000 +00000000000000000000000000000000 0.29999999999999998889776975374843459576368331909179687500000000000000 +00000000000000000000000000000000 0.30000000000000004440892098500626161694526672363281250000000000000000 +00000000000000000000000000000000` [download]	[reply] [d/l] [select]

In Section Meditations