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
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
