Hacker News Reader

Does this mean simple for-loops in R will be faster? Currently R can be order(s) of magnitude slower than MATLAB / Python. Does one have to do something specific to use the JIT or is it automagic?

View on HN

Speed differences of that magnitude are largely due to bad interactions with R's data allocation model. R for loops are not that slow if there is no allocation going on, and there are ways to organize the allocation ahead of time: http://www.win-vector.com/blog/2015/07/efficient-accumulatio...

So my guess is: no the JIT won't help for ultra slow code as in my experience that isn't the problem in that case.

Really nice article, shows why writing fast R code is such a pain. I had no idea that data frame is that much worse than data table.

While this is a great article, it also transmits some truth about R: writing anything fast is a pain in the ass. Especially frustrating when you hear those R "experts" repeating "use apply" functions. The truth is, the design of the language requires you to actually learn some dirty details and NOT solely rely on the idiomatic syntax.

I like R quite much, but this is a sign of a bad language design.

Reply on news.ycombinator.com