"Tore Haug-Warberg" <haugwarb / chembio.ntnu.no> schrieb im Newsbeitrag
news:4.3.2.7.0.20030624115646.00be2100 / pop.chembio.ntnu.no...
>
> At 18:41 24.06.03 +0900, Tim Bates wrote:
> >On Tue, Jun 24, 2003 at 05:54:42PM +0900, Holden Glova wrote:
> > > Actually funny you ask because I can't seem to make the float class
> > provide
> > > rounding errors. I have been searching on the web but can't seem to
> > make Ruby
> > > give me rounding errors. I was under the impressions that floats
where
> > always
> > > inaccurate, is this not the case in Ruby?
>
> Interesting. I have also noted this and figured out that maybe Matz has
> been clever enough to design the input/output routines of float numbers
> such that the process is reversible (I mean from float to string and
back).
> However, I did not trust this and besides it is the issue of
reversibility
> from string to float and back (more than, say, 17 digits in the original
> string).

There are rounding errors:

irb(main):013:0> x=1.0
1.0
irb(main):014:0> e=1e-100
1e-100
irb(main):015:0> y=x+e
1.0
irb(main):016:0> y==x
true
irb(main):017:0>

.... but obviously y>x

    robert