This is a summary of my experience mentoring and administering the R project in Google Summer of Code 2017.
We had 35 projects funded at the beginning of summer. Two of these projects failed; in one case the student disappeared, in another case the student decided to work on another full-time job instead. This year the pass rate for the R project is 33/35 = 94.3%, which is higher than the overall average of 85.6% from GSOC2016. I mentored three of the projects that passed.
Marlin Na wrote the TnT package for rendering interactive genome browsers in R. This project was his idea, and he was very self motivated. We did not have skype calls, but he provided very detailed email updates all throughout GSOC. He wrote a blog post about his experience, and plans to submit the package to Bioconductor.
Rover Van implemented speed optimizations for the iregnet package which provides a machine learning algorithm for regression with censored outputs. His commits for GSOC2017 are summarized on PR59, and his blog post provides a nice description of his overall experience. Although the code is still not as fast as glmnet, his benchmarks clearly show that the code is now faster than it was before GSOC.
It was my second year working on the interactive grammar of graphics with Faizan Khan. As summarized on his blog post, his GSOC project was a very ambitious rewrite of the original animint package. This resulted in animint2, which supports an simpler syntax for defining interactivity using parameters rather than aesthetics. It also has dropped the dependency on ggplot2; we now instead use the ggplot2Animint fork, which will be more stable and easier to maintain (no need to provide updates every time the ggplot2 developers make backwards incompatible changes).
Overall it was another very successful year for the R project in Google Summer of Code, and I look forward to participating again next year. In fact, I have already setup a wiki with a few project ideas for R in GSOC2018.