StatsRhian
commented
3 months ago Dropping in my notes. Might try and quickly write this up in the mid-sprint week?
Why R packages
- Questions
- Have you made an R package?
- When should you make an R package?
- Benefits
- Reusability across projects
- Easy to use documentation
- Testing
- Dependency management
- Myth busting
- Doesn’t have to be on CRAN
- Doesn’t have to have hundreds of functions
- Doesn’t have to be useful for anyone else
- It doesn’t have to do anything fancy
- It’s hard to make an R package
- Talk through the structure
- Show {usethis} for quick R package
- Nice-to-have’s
- pkgdown
- GitHub Actions
- CRAN
- Hex stickers
Documentation
- Demystify
- “meta context”
- Vignettes
Our R packages
Why?
- Share with other analysts (inside & outside of SU)
- Easier to debug outside of Outputs app
- Easier to develop new functions (don’t need a shiny expert)
Why now?
- Finalised reports are coming
- Analysts are starting to duplicate work
- About to re-visit Waterfall / Tornado
What does done look like?
- Package that can take JSON and produce waterfall/beeswarm etc
Why did we start
- We repeat repeat repeat ourselves (outputs app)
- Changing code once to effect lots of things (update calc - improve docs)
What did we think about
- Where should stuff live?
- Who is going to use it?
- Why two packages?
- Azure is used in NHP - but also useful for others
su.azure
- Wrapping R packages is okay
- Fragmenting - teasing out purpose
nhp.results
What the next for these packages
- Our vision?
- NHP outputs (just shiny)
- Fix as we touch
- Little bits at a time
- NHP analysis (can stay)
- Do you have repeated code?
- How stable are they going to be?
- We’re trying to be aware
- Style book (Hadley)
- How do I use it?
- GitHub install / PAT
- Your country needs you (Peter)
- Functionality
- Issues
matt-dray
On 2024-07-12. Why might you need to make one, package-dev flow, etc.