Hi community, I already participated in several discussions with some of you regarding adding support to R on Marvin. Allowing users to write engines in R would be a great step for Marvin since that's the primary language for many statisticians, physicists and math people.
One of the points that always arise during those conversations is the fact that R is not fast nor scalable, and it would be hard to either make it process large chunks of data for training or make it support high throughput and have low latency for predictions. I want to start a thread on whether we should invest more time trying to add R support to Marvin or not. I invite you all to read about this related work: https://medium.com/graalvm/faster-r-with-fastr-4b8db0e0dceb I am looking forward to hearing opinions and ideas. Thanks, Lucas
