How to Pretend Typing Super Fast in RStudio

A blog post written during my presentation at rstudio::conf 2018

Yihui Xie 2018-02-02

Right now I’m giving a talk at the RStudio Conference in San Diego. In my talk, I told the audience that I was going to write a blog post at the podium during my presentation. I only had 20 minutes in total for my presentation. Apparently it was impossible to type out a nontrivial post. But I want to show them my workflow of writing and publishing blog posts, so I cheated using the rstudioapi package to automatically “type” the post in RStudio for me.

The rstudioapi package

The key idea is to use the function rstudioapi::insertText() to insert text in the RStudio source editor, and the editor context can be obtained from rstudioapi::getSourceEditorContext(). The rest of things to do is a plain for loop, in which you insert characters one by one, and use Sys.sleep() to pause occasionally (and randomly), otherwise it will be obviously bogus.

Below is a prototype of the function to automatically “type” a character vector into your RStudio source editor:

rstudio_type = function(x) {
  ctx = rstudioapi::getSourceEditorContext()
  x = paste(x, collapse = '\n')  # concatenate by \n
  x = unlist(strsplit(x, ''))    # split into individual characters
  for (i in x) {
    rstudioapi::insertText(text = i, id = ctx$id)
    Sys.sleep(rbeta(1, 1, 9)/10)  # make delay time right-skewed


I polished up the code and included it in the xfun package as the function xfun::rstudio_type(). I added a few arguments to the function to make it feel more like a real human typing. In particular, the mistake argument controls how frequently you make mistakes (insert random strings that will be deleted shortly). Currently you have to install the development version of xfun to use this function:


How did it work when I gave the talk?

I typed go in my R console, and hoped nobody saw it. The object go was created from a delayed assignment:

delayedAssign('go', {
    pause = function() rbeta(1, 1, 9)/60, mistake = .001, save = .005

You may ask why not define a normal function, like go = function() { ... }? I didn’t want to type go() in my console, because that would look suspicious to the audience whose eyes are really good, and they might know I called a function to do something. If I just type go, even if they saw it, I don’t think they would so quickly think of delayed assignments in R, and they might think I was trying to print a certain object in R, but didn’t see its value for some reason after I press Enter.

Update on 2018-03-04

The video of the talk is freely available on the RStudio website.

Create and maintain websites with R Markdown and blogdown – RStudio

Appendix: Does yolo: true work in blogdown?

Brandon Hurr asked me this question on Twitter a couple of days ago. Well, yolo: true in the xaringan package is essentially the Mustache Karl picture inserted on random slides. I can certainly make it work for blogdown (i.e., insert Karl’s picture randomly in your blog posts), but I’d prefer not to, since xaringan is pretty much a personal package, and blogdown is not. I don’t wish to push it too far.

That said, you have full freedom on your personal blog, so let met tell you the secret weapon. There is an unexported function yolo() in xaringan, and you can call it in an R code chunk, e.g.,


It should give you what you want:

Moustache Karl (TM)