search

statist7 / LMSgrowth2

MIT License
2 stars 2 forks source link

Set up environment for development #14

Open tamuri opened 5 years ago

tamuri commented 5 years ago

Includes guide for developers on setting up for development

tamuri commented 5 years ago

A possible alternative to full containerisation (at least for development) is to use Packrat to manage the R dependencies (it's like Python's virtualenv). Might be preferable to full containerisation (especially as we're only using the R ecosystem)

Answer about how git and packrat work together.

statist7 commented 5 years ago

Happy to go with what's simplest.

tamuri commented 5 years ago

Just noting here FYI: came across an issue setting up packrat on MacOS. install.packages on MacOS by default installs binaries and does not build from source. This means it skips some of the dependencies listed in "LinkingTo" (e.g. dplyr LinkingTo plogr). packrat::snapshot() complains about missing packages.

After installing the packages packrat complains about (BH, plogr), you can packrat::snapshot(). I don't think it'll be a problem.

Also https://github.com/rstudio/packrat/issues/541

tamuri commented 5 years ago

I'll mention here: I did a bit of digging into R environments for multiple versions of R and conda feels like the only pain-free way to do that. I would like to add this to dev setup because packrat does not like it if you're installing packages on an old version of R