how about is_evenly_spaced <- function(x,...) all.equal(diff(sort(x)),...)
(use ellipsis to set tolerance if necessary) Op vr 31 aug. 2018 om 15:46 schreef Emil Bode <[email protected]>: > 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 > (<[email protected]>) 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]] > > > > ______________________________________________ > > [email protected] mailing list > > https://stat.ethz.ch/mailman/listinfo/r-devel > > > > -- > Iñaki Ucar > > ______________________________________________ > [email protected] mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel > > > ______________________________________________ > [email protected] mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel > [[alternative HTML version deleted]] ______________________________________________ [email protected] mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
