[Bioc-devel] Use set.seed inside function

[Henrik Bengtsson]

The easiest is to use withr::with_seed(), e.g.

> withr::with_seed(seed = 42L, randomcoloR::distinctColorPalette(6))
[1] "#A0E1BC" "#B8E363" "#D686BE" "#DEA97F" "#B15CD8" "#A2B9D5"
> withr::with_seed(seed = 42L, randomcoloR::distinctColorPalette(6))
[1] "#A0E1BC" "#B8E363" "#D686BE" "#DEA97F" "#B15CD8" "#A2B9D5"

It works by undoing globalenv()$.Random.seed after the random number
generator has updated.  If you want to roll your own version of this,
you need to make sure to handle the special case when there is no
pre-existing .Random.seed in globalenv().

Regarding packages and functions changing the random seed via a
set.seed() [without undoing it]: this should *never* be done, because
it will wreak havoc on a analyses and studies that rely on random
numbers.  My rule of thumb: only the end-user should be allowed to use
set.seed(), which should typically be done at the top of their R
scripts.

/Henrik

On Mon, Nov 29, 2021 at 1:23 PM Meng Chen <mengchen18 using gmail.com> wrote:
>
> Thanks. I think it may work in theory, generating "enough" distinct colors
> is fairly easy. Then the problem will be how to find a subset of colors of
> size n, and the selected colors are still most distinguishable. I think I
> will do this with my eyes if no other methods, a tedious job.
> But at least for my curiosity, I still want to know if there are other ways
> to achieve this. I feel like 80% of people who use the distinctColorPallete
> function actually don't need the "random" feature :) Thanks.
>
> On Mon, Nov 29, 2021 at 9:39 PM James W. MacDonald <jmacdon using uw.edu> wrote:
>
> > It appears that you don't actually want random colors, but instead you
> > want the same colors each time. Why not just generate the vector of 'random
> > distinct colors' one time and save the vector of colors?
> >
> > -----Original Message-----
> > From: Bioc-devel <bioc-devel-bounces using r-project.org> On Behalf Of Meng Chen
> > Sent: Monday, November 29, 2021 3:21 PM
> > To: bioc-devel using r-project.org
> > Subject: [Bioc-devel] Use set.seed inside function
> >
> > Dear BioC team and developers,
> >
> > I am using BiocCheck to check my package, it returns a warning:
> > " Remove set.seed usage in R code"
> >
> > I am using "set.seed" inside my functions, before calling function
> > distinctColorPalette (randomcoloR package) in order to generate
> > reproducible "random distinct colors". So what would be the best practice
> > to solve this warning? I think 1. use set.seed and don't change anything.
> > 2. use the set.seed function, but include something like below inside the
> > function *gl.seed <- .Random.seed* *on.exit(assign(".Random.seed", gl.seed,
> > envir = .GlobalEnv))* 3. use some other functions for the purpose
> >
> > Any suggestions will be appreciated. Thanks.
> > --
> > Best Regards,
> > Chen
> >
> >         [[alternative HTML version deleted]]
> >
> > _______________________________________________
> > Bioc-devel using r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/bioc-devel
> >
>
>
> --
> Best Regards,
> Chen
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> Bioc-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel