library(methods): Human Time vs Computer Time

(This article was originally published at Yihui's Blog on Yihui Xie | 谢益辉, and syndicated at StatsBlogs.)

There is a forever confusing thing when you run the command Rscript: it does not attach the methods package by default. It is confusing because when you start an R session interactively, methods is loaded and attached by default. When you run the command R, it is also attached. It is just Rscript that does not load methods.

This confuses users over, and over, and over again, especially if you love or touch S4 (directly or indirectly). You will surely see questions like this one in the future unless the base R developers decide to change Rscript. It is said that the reason for not loading the methods is because it is slow, “in history”, to load this package.

Okay, let’s see how terribly “slow” it is today. On my MacBook Pro (I should run this multiple times but I think 0.04 should be in the ballpark):

$ Rscript -e "system.time(library(methods))"
   user  system elapsed 
  0.035   0.003   0.039

Is 0.04 seconds worth saving? I doubt it. Should it be considered as terribly slow? Here is the loading time of the stats package:

$ R_DEFAULT_PACKAGES=base Rscript -e "system.time(library(stats))"
   user  system elapsed 
  0.041   0.013   0.055 

Well, that is not a fair comparison, because stats depends on other base packages like grDevices, so here is the result for grDevices:

$ R_DEFAULT_PACKAGES=base Rscript -e "system.time(library(grDevices))"
   user  system elapsed 
  0.019   0.007   0.029 

I’d say methods is relatively slow to load, but it is not too bad. Given the endless confusion caused by Rscript, I don’t think it is worth excluding it by default. In this case, we have wasted way too much human time on explaining the confusion to save a tiny bit computer time. If users really care about this 0.04 seconds, they should set the environment variable R_DEFAULT_PACKAGES by themselves, and I doubt anyone in the world would do it at all.

It is sad to see things that exist for historical reasons are so hard to change.

P.S. Although S4 is innocent in this case, this is a minor reason why I never liked S4.

Update on 2017/01/06: Gabor discovered that Luke Tierney finally made Rscript consistent with R by default, i.e., methods will be attached in a future version of Rscript. It seems the long-lasting confusion will be ended in 2018.



Please comment on the article here: Yihui's Blog on Yihui Xie | 谢益辉


Subscribe

Email:

  Subscribe