R sample returns always the same sample

[piccolbo]

it's supposed to be uniform sampling with or without replacement.

[piccolbo]

So the problem here is that getRandomSample in Java and scala requires a seed, like takeSample in spark. But that's not how sample should work because it's supposed to provide independence between calls when seed is NOT provided. Furthermore the R function provides a default seed = 123 hence the repeated results. I have no doubts this should be randomized, even if sample doesn't have a method for data frames, it's clear how it works on vectors. The question is, should we fix this also in Java/scala or not? When the seed is a mandatory argument, this is not a bug, maybe a missing feature. This poses the larger question of how much we want to align the APIs between languages.

[piccolbo]

Any comments @nhanitvn ?

[piccolbo]

please @nhanitvn review

[ctn]

@nhanitvn please review & close as appropriate.