Today I struggled for hours to understand some unexpected package test results. It turned out that this is because package "parallel", buried deep in my dependencies, calls runif() during it's initialization and in this way changes the random number sequence. This seems to be a part of a more general question--which kind of functions can we trust if we want to preserve random numbers, and the rest of the environment, like graphical parameters?
One could put all the burden on the end user--never trust any of the calls you make, it is your responsibility to save the RNG state if you wish. However, this might be very cumbersome if you have to do this around all cat(), print() and similar calls that most likely will never tinker with random numbers. For instance, can I be sure that set.seed(0); print(runif(1)); print(rnorm(1)) will always print the same numbers, also in the future version of R? There are also commands that we definitely expect to change the RNG state, like random numbers, bootstrapping, etc. I just did not expect that library() is one of them... I think it would be nice to have some sort of guidelines here. The only related sentence I am aware of is in the CRAN policy: - Packages should not modify the global environment (user’s workspace). Any thoughts? Cheers, Ott -- View this message in context: http://r.789695.n4.nabble.com/Which-function-can-change-RNG-state-tp4702935.html Sent from the R devel mailing list archive at Nabble.com. ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
