The lazy numbers in R: correction

Posted on November 15, 2022 by Stéphane Laurent
Tags: R, C++, Rcpp, maths

Because of a change I did in the lazyNumbers package, I have to post a correction to my previous post.

The as.double function, called on a lazy number, was not stable. Now it is. In the previous post, the following equality was true:

x <- 1 - lazynb(7) * 0.1
as.double(x) == 0.3
## [1] FALSE

It is not true anymore. This is expected actually. Indeed, the double numbers 0.1 and 0.3 do not exactly represent the numbers \(0.1\) and \(0.3\):

print(0.1, digits = 17L)
## [1] 0.10000000000000001
print(0.3, digits = 17L)
## [1] 0.29999999999999999

The double representation of whole numbers is exact. The following equality is true:

x <- 1 - lazynb(7) / 10
as.double(x) == 0.3
## [1] TRUE

No other change, and nothing else to correct. It is time to submit the package to CRAN. See the Github repository for another short presentation of this package.