How to Become an Efficient and Collaborative R Programmer | Statistics, R, Graphics and Fun | Yihui Xie | /en/2011/12/how-to-become-an-efficient-and-collaborative-r-programmer/

yihui 2022-12-16 21:03:06

https://yihui.org/en/2011/12/how-to-become-an-efficient-and-collaborative-r-programmer/

4 Comments

giscus-bot 2022-12-16 21:03:07

Guest *Songpants* @ 2011-12-13 13:43:08 originally posted:

Nice post!

giscus-bot 2022-12-16 21:03:08

Guest *Barry Rowlingson* @ 2011-12-13 17:58:17 originally posted:

Thanks for the namecheck!

The only reason for using R-Forge at the moment is getting the Windows builds done for tricky packages, and being able to install directly from it via install.packages. That's two reasons I suppose...

I wonder if you could set up an automatic system to do package builds (and somehow do them for architectures you don't have handy, perhaps that win-builder site) and upload them to a github download site. Then how could you make that work with install.packages....

Hmmmm

yihui 2022-12-16 21:03:10

Hadley has install_github() in his devtools package (requires Rtools) to install from GitHub directly, and this package also allows you to upload to that win-builder site using R functions.

I do not care much about how Windows users install packages under development on GitHub using install.packages(); after the package is uploaded to CRAN, everything will be fine.

Originally posted on 2011-12-13 22:58:46

giscus-bot 2022-12-16 21:03:12

Guest *Charlie Sharpsteen* @ 2011-12-19 04:33:23 originally posted:

I keep putting up with R-Forge because it is the easiest way to do incremental development builds between releases. This is useful for a quick sanity check to make sure the package isn't broken on other platforms. R-Forge also provides an easy way for my users to grab development builds using install.packages. This way, they don't have to install an R development environment in order to build from source.

However, it seems that R-Forge is always half broken. For many platforms, the automatic builds and tests are listed as "down for maintenance" and have been so for more than a year.

What we really need is a Continuous Integration and packaging platform for R. Something like Travis-CI:

http://travis-ci.org

Then developers could host code wherever they please and still have a services that could run checks across multiple platforms and create binary packages.

yihui 2022-12-16 21:03:13

I just tried Cameron's approach to commit the git repository to r-forge; he said the instructions were from you, so thanks a lot! I will see how the half broken r-forge performs.

Originally posted on 2012-03-14 06:35:17

giscus-bot 2022-12-16 21:03:08

Guest *Mike* @ 2011-12-13 20:28:23 originally posted:

Would you mind explaining how you use Github with your R programming? I've yet to find an awesome workflow for how to incorporate the two. Ruby - yes. Python - yes. R - no.

Is there a package I should download that allows me to push to github?

Thanks

yihui 2022-12-16 21:03:11

To use GitHub, you need to learn some basic GIT commands, and yes, you have to download/install GIT packages (see http://help.github.com/).

For development in R, I'm basically following the steps of Hadley: https://github.com/hadley/devtools/wiki The outline is there and quite a few details still need to be added (one thing missing there is the package vignette, which I believe is easy for me)

Originally posted on 2011-12-14 01:20:33

giscus-bot 2022-12-16 21:03:09

Guest *Gang Chen* @ 2011-12-18 20:03:52 originally posted:

It seems that a GitHub account with lots of commit records is required for all coders. I have a GitHub account, but I am using Google Codes. Maybe, I should put my next project to GitHub.

Some people are working on a clone of GitHub, TeamHost.org, to provide fast service for Chinese developer. But it is still in very early stage.