Impute KNN sets seed

[const-ae]

Hi Laurent,

first thanks for the great package, it makes dealing with MS data much simpler.

I was just exploring the different methods for imputing missing values, when I discovered that to my surprise impute(msnset, "knn") fixes the seed value.

suppressPackageStartupMessages(library(MSnbase))
data(msnset)
# Other methods show the expected behavior
for(i in 1:10){
  tmp <- MSnbase::impute(msnset, "zero")
  print(rnorm(1))
}
#> [1] 2.077938
#> [1] -0.3538495
#> [1] 1.269365
#> [1] 0.1198152
#> [1] -0.07994367
#> [1] -1.107776
#> [1] 0.5004011
#> [1] -0.7553048
#> [1] -1.225024
#> [1] 0.2958373

# Fixes the seed, so no more randomness afterwards
for(i in 1:10){
  tmp <- MSnbase::impute(msnset, "knn")
  print(rnorm(1))
}
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527
#> [1] -0.668527

^{Created on 2018-11-15 by the reprex package (v0.2.1)}

I saw that internally it is actually not your package that calls set.seed(), but the impute::impute.knn() method, but the method returns the state of the random number generator. So I was wondering if you could reset the rng state after the method call? If you are currently busy I could also try to make a small PR.

Best, Constantin

[lgatto]

Thank you for this, could catch. I wouldn't mind a PR.