Le 31/08/2018 à 16:25, Mark van der Loo a écrit :
Ah, my bad, you're right of course.
sum(abs(diff(diff( sort(x))))) < eps
for some reasonable eps then, would do as a oneliner, or
all(abs(diff(diff(sort(x)))) < eps)
or
max(abs(diff(diff(sort(x))))) < eps
Or with only four function calls:
diff(range(diff(sort(x)))) < eps
Serguei.
-Mark
Op vr 31 aug. 2018 om 16:14 schreef Iñaki Ucar <iu...@fedoraproject.org>:
El vie., 31 ago. 2018 a las 16:00, Mark van der Loo
(<mark.vander...@gmail.com>) escribió:
how about
is_evenly_spaced <- function(x,...) all.equal(diff(sort(x)),...)
This doesn't work, because
1. all.equal does *not* return FALSE. Use of isTRUE or identical(.,
TRUE) is required if you want a boolean.
2. all.equal compares two objects, not elements in a vector.
Iñaki
(use ellipsis to set tolerance if necessary)
Op vr 31 aug. 2018 om 15:46 schreef Emil Bode <emil.b...@dans.knaw.nl>:
Agreed that's it's rounding error, and all.equal would be the way to go.
I wouldn't call it a bug, it's simply part of working with floating
point numbers, any language has the same issue.
And while we're at it, I think the function can be a lot shorter:
.is_continous_evenly_spaced <- function(n){
length(n)>1 && isTRUE(all.equal(n[order(n)], seq(from=min(n),
to=max(n), length.out = length(n))))
}
Cheers, Emil
El vie., 31 ago. 2018 a las 15:10, Felix Ernst
(<felix.gm.er...@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@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
--
Iñaki Ucar
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Iñaki Ucar
[[alternative HTML version deleted]]
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
--
Serguei Sokol
Ingenieur de recherche INRA
Cellule mathématiques
LISBP, INSA/INRA UMR 792, INSA/CNRS UMR 5504
135 Avenue de Rangueil
31077 Toulouse Cedex 04
tel: +33 5 62 25 01 27
email: so...@insa-toulouse.fr
http://www.lisbp.fr
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
