Yihui Xie

Once in a Blue Moon, I Reached "Issue Zero" in a Github Repo

The new CRAN release of the rticles package

Yihui Xie / 2018-07-06


“Inbox Zero” (no emails in the inbox) may never be possible to me again, but today I reached a goal of “Issue Zero” for the Github repo rstudio/rticles. I have to say this is so rare. In fact, it is the very first time I have been able to close all issues in a Github repo.

no more rticles issues

Answer all questions and kill all bugs

Last week we were reminded of the fact that the rticles package had not been updated on CRAN for more than a year. Its original maintainer (JJ) asked me for help, so I spent the last week and this week on the new CRAN release. Initially I only planned to take a look at existing pull requests, which contained new article formats contributed from other users. After reviewing all pull requests, I also looked at the issues. I knew there was an annoying long-standing bug (you could not use $ in JSS or R Journal articles), which I really hoped to fix, because it has affected multiple users (#6, #10, #49, #70, #132, #149, and rstudio/rmarkdown#398).

Once I started working on Github issues, I thought it might be good to take advantage of the inertia, which is often hard to obtain. Finally I answered all questions and fixed all bugs reported in these issues. During this process, I realized that we didn’t even support bibliography for The R Journal articles. That was such a shame. How could we have provided such a half-baked product? It revealed the fact that rticles maintainers at RStudio didn’t really write journal articles with the rticles package (i.e., we don’t eat our own rticles dog food), which means we are probably not the most appropriate people to work on these article formats. We absolutely need more contribution from real authors and LaTeX experts. On the other hand, this also means rticles users may have better luck when asking questions in other public forums such as Stack Overflow or RStudio Community instead of Github issues, because we maintainers are not experts of these article formats, and I don’t want you to wait for four years for me to answer your question.

No more looking for missing LaTeX packages

The best thing I did for this release was enabling Travis CI for this repo (thanks to Zhian N. Kamvar). Yes, it was so late, and such a big shame. We should have enabled Travis CI years ago. Previously JJ was just trying to manually test everything locally on his computer when new pull requests came in.

One nice thing about the Travis setup is that it uses (and caches) TinyTeX, so that any missing standard LaTeX packages required by the article formats can be automatically installed on Travis (and cached). No sudo is required.

cache:
  packages: true
  directories:
    - $HOME/.TinyTeX
latex: false

env:
  global:
    - PATH=$HOME/bin:$PATH

before_install:
  - wget -qO- https://github.com/yihui/tinytex/raw/master/tools/download-travis-linux.sh | sh

At this moment, the setup for R on Travis adds symlinks of TeX Live executables to /usr/local/bin, which prevents us from installing other executables (such as epstopdf). That is why I disabled the default LaTeX installation latex: false.

Hi, Journal of Statistical Software

Early last year, the editor of JSS said the JSS article format in rticles was “unacceptable”. I was sad to hear that. I’m always sad to hear formatting issues could be real and critical issues. Anyway, after some communication with the editor, we happily and easily resolved the main issue (bibliography should have been generated via natbib). So it was actually not too bad. After fixing the dollar sign issue mentioned before, and the issue of authors on multiple lines (#100), personally I feel the JSS article format in rticles is pretty good now. The new version of rticles (v0.5) has been released on CRAN. Let’s see if we can survive the journal of the highest LaTeX standard in the world this time.

Am I awesome enough now?