Hacker News Reader

Comment by minimaxir | Hacker News Reader

R Markdown is very powerful. That being said, I still find myself using Jupyter notebooks instead due to first-class language kernel support of non-R languages (which was not the case at Rmarkdown's launch, when Jupyter was still IPython), and GitHub integration for rendering notebooks.

The dynamic aspect of Rmarkdown is done through HTMLWidgets, which also works with Jupyter notebooks too. (But not via the GitHub integration mentioned earlier.)

View on HN

Have you checked out R Markdown notebooks?

http://rmarkdown.rstudio.com/r_notebooks.html

Huh, no I haven't. That feature seems fleshed out so I'll take a look!

My understanding is that full notebook support will be released in the next version of RStudio. It's available in the daily/preview builds of Rstudio, which have been fairly stable for me over the last week.

https://www.rstudio.com/products/rstudio/download/preview/

I didn't see it in the preview images, but will the notebooks have the usual Rstudio features of dataset previews and file explorer?

Yeah - it's basically the same workflow as Rmarkdown documents now but with inline plotting, instead of plotting to the 'plot' window (or whatever other attached graphics devices you have).

A couple of issues I have had is that it only suppoets the R version in the Anaconda repository and thus you have to reinstall packages. It also means you have to point directly to a CRAN repository to get packages.

Are you referring to RStudio or Jupyter?

Jupyter

I've almost fully switched from Jupyter to RMarkdown after discovering RMarkdown supports Python. And about publishing / rendering, do you not like RPubs?

As well as a number of other rather useful languages [1].

Now days I use RMarkdown even for experiments because it's nice to document my line of thinking as I write the code and get a nicely formatted output when I'm finished.

[1] http://yihui.name/knitr/demo/engines/

I saw the Python examples provided awhile ago and they looked hacky. It might be worth taking another look, though.

I prefer make the notebooks portable, which I can't do if they are published on Rpubs.

I still feel the REPL-like environment in Jupyter is more suitable for building an idea, while RMarkdown is better for publishing it.

Do HTMLWidgets really work with Jupyter? I saw there is an issue but it's still open: https://github.com/ramnathv/htmlwidgets/issues/144

Or are you referring to ipywidgets, not literally HTMLWidgets?

Hmm. I know plotly did as the documentation explicitly covers that use case. But the comments in the linked issue imply it is a hack.

Reply on news.ycombinator.com