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