# [Rd] compairing doubles

Iñaki Ucar iuc@r @ending from fedor@project@org
Fri Aug 31 15:49:00 CEST 2018

```FYI, more fun with floats:

> 0.1+0.1==0.2
[1] TRUE
> 0.1+0.1+0.1+0.1==0.4
[1] TRUE
> 0.1+0.1+0.1==0.3
[1] FALSE
> 0.1+0.1+0.1==0.1*3
[1] TRUE
> 0.3==0.1*3
[1] FALSE

¯\_(ツ)_/¯

But this is not R's fault. See: https://0.30000000000000004.com

Iñaki

El vie., 31 ago. 2018 a las 15:36, Iñaki Ucar
(<iucar using fedoraproject.org>) escribió:
>
> El vie., 31 ago. 2018 a las 15:10, Felix Ernst
> (<felix.gm.ernst using outlook.com>) escribió:
> >
> > Dear all,
> >
> > I a bit unsure, whether this qualifies as a bug, but it is definitly a strange behaviour. That why I wanted to discuss it.
> >
> > With the following function, I want to test for evenly space numbers, starting from anywhere.
> >
> > .is_continous_evenly_spaced <- function(n){
> >   if(length(n) < 2) return(FALSE)
> >   n <- n[order(n)]
> >   n <- n - min(n)
> >   step <- n[2] - n[1]
> >   test <- seq(from = min(n), to = max(n), by = step)
> >   if(length(n) == length(test) &&
> >      all(n == test)){
> >     return(TRUE)
> >   }
> >   return(FALSE)
> > }
> >
> > > .is_continous_evenly_spaced(c(1,2,3,4))
> > [1] TRUE
> > > .is_continous_evenly_spaced(c(1,3,4,5))
> > [1] FALSE
> > > .is_continous_evenly_spaced(c(1,1.1,1.2,1.3))
> > [1] FALSE
> >
> > I expect the result for 1 and 2, but not for 3. Upon Investigation it turns out, that n == test is TRUE for every pair, but not for the pair of 0.2.
> >
> > The types reported are always double, however n[2] == 0.1 reports FALSE as well.
> >
> > The whole problem is solved by switching from all(n == test) to all(as.character(n) == as.character(test)). However that is weird, isn’t it?
> >
> > Does this work as intended? Thanks for any help, advise and suggestions in advance.
>
> I guess this has something to do with how the sequence is built and
> the inherent error of floating point arithmetic. In fact, if you
> return test minus n, you'll get:
>
> [1] 0.000000e+00 0.000000e+00 2.220446e-16 0.000000e+00
>
> and the error gets bigger when you continue the sequence; e.g., this
> is for c(1, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7):
>
> [1] 0.000000e+00 0.000000e+00 2.220446e-16 2.220446e-16 4.440892e-16
> [6] 4.440892e-16 4.440892e-16 0.000000e+00
>
> So, independently of this is considered a bug or not, instead of
>
> length(n) == length(test) && all(n == test)
>
> I would use the following condition:
>
> isTRUE(all.equal(n, test))
>
> Iñaki
>
> >
> > Best regards,
> > Felix
> >
> >
> >         [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-devel using r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
>
> --
> Iñaki Ucar

--
Iñaki Ucar

```

More information about the R-devel mailing list