best-practices.Rmd

# Best practices

"Each of you is perfect the way you are ... and you can use a little improvement." 
--- Shunryu Suzuki

## Priority list

### Personal packages

Use RStudio. 

Use RStudio Projects. Learn the keyboard shortscuts, especially Cmd + Shift + L for reloading all your code. Master the code development workflow of edit, reload and interactively test.

Use version control.

Use formal tests 

Learn C++

### Shared packages

use roxygen

use github

use R CMD check

Release to CRAN.

## Development cycles

* fixing a bug/implementing a feature: edit -> load_all() -> interactive exploration -> write tests

* push to github: as above + check()

* use a branch: basically I only do this when I'm working on a package with someone else and either want to get their feedback before merging, or we're working simultaneously and I don't want to tread on their toes.

* release new version to CRAN: This tends to be mostly driven by external factors. Often I'll have some big picture plans for high profile packages (e.g. for each dplyr release I've had a pretty good feel for what the headline features will be), but for most of my smaller packages it's more driven by mental exhaustion from other projects, so I want something simple to do.  Or github issues accumulate for several years, and then finally it tips the balance and I fix them all. Batching is super important so that I can avoid expensive context-switching