R package development

# R package development
## - creating and distributing -
### Athanasia Monika Mowinckel
### 20.01.2022

---

background-image: url("https://drmowinckels.io/about/profile.png")
background-position: right bottom
background-size: auto 100%
class: middle, dark

[<i class="fa fa-twitter fa-2x" aria-hidden="true"></i> @DrMowinckels](https://twitter.com/DrMowinckels)   
[<i class="fa fa-github fa-2x" aria-hidden="true"></i> Athanasiamo](https://github.com/Athanasiamo)  
[<i class="fa fa-globe fa-2x" aria-hidden="true"></i> drmowinckels.io/](https://drmowinckels.io/)

- Staff scientist / <br>
   Research Software Engineer  
- PhD in cognitive psychology  
- Software Carpentry & <br>
  RStudio tidyverse Instructor   
- R-Ladies Oslo & <br>
  [R-Ladies Global team](www.rladies.org)
]

---

---
class: dark

# Why should you bother with R?

- None of your colleagues use R

- You are happy with the software you know

- You can create transparent and reproducible workflows in the tools you already know

]

- help colleagues start working more transparently and reproducibly

- take advantage of the huge repertoire of packages to aid in almost any statistical endeavour

]

???
And R can these days do much more than statistics, though it is fast becoming a very popular tool with data scientist.
R is often quite popular in the social sciences, and in my opinion better options that Stata or SPSS, which are usually taught.

---
class: dark

# It sounds like lots of work...

- does require initial efforts to get into a package format
    - utilise all the packages available to help you create, test, distribute and document your package

- well-documented and stable functions reduce long-term maintenance and need of explanations to colleagues already familiar with working in R

- unit tests of package functionality can alert you to changes in R or package dependencies that break core package functionality
    - or to protect from future you making breaking changes

- [oRganization: How to make internal R packages part of your team](https://www.rstudio.com/resources/rstudioglobal-2021/organization-how-to-make-internal-r-packages-part-of-your-team/) - Emily Reiderer at Rstudio::Global 2021

---

# Talk outline

- Identify and assess staff needs
- Key helper packages
- Take a look at an in-house package
- Distributing packages
   - locally
   - online

---

# Identify and assess staff needs

- What are the things you observe them doing in very manual and tedious ways?

- What operations require customised handling for your data/tasks?
]

- What is the user level proficiency?

- What operating systems are people working on?

- How are people used to working in R?
]

---
class: dark
background-image: url("https://upload.wikimedia.org/wikipedia/commons/d/d0/RStudio_logo_flat.svg"), url("https://upload.wikimedia.org/wikipedia/commons/9/9a/Visual_Studio_Code_1.35_icon.svg")
background-position: 15%, 85%
background-size: 40%, 20%

# OK, cool, cool. How do I develop?

???
you have mainly two developer IDE's to choose from: RStudio and vscode.
I use RStudio, it has a lot of functionality that helps develop packages and functions efficiently, and start checks and builds. 
I've never personally used vscode for R, but have colleagues who have for general R use, but not for package dev, so I can't really say anything for or against it.
I'll be using RStudio, and so you can see the process and how the IDE helps me be more efficient.

---

# Key helper packages

- [usethis](https://usethis.r-lib.org/) to help set-up pacakge
  - licences, github actions, online documentation
- [roxygen2](https://roxygen2.r-lib.org/) inline function documentation
- [devtools](https://devtools.r-lib.org/) package loading, checking, building
- [testthat](https://testthat.r-lib.org/) unit testing
- [rmarkdown](https://github.com/rstudio/rmarkdown) package documentation as vignettes/tutorials

---
class: inverse, middle, center

# Example package
## [Questionnaires](https://lcbc-uio.github.io/Questionnaires/index.html)

---

# Distributing packages

- CRAN
  - requires diligence and patience
  - [follow steps carefully](https://github.com/ThinkR-open/prepare-for-cran)
- GitHub
  - `remotes::install_github("user/repo")`
  - will not compile for all users, can be tricky on networked windows
]

- [R-universe](https://ropensci.org/r-universe/) by [rOpenSci](https://ropensci.org/)
  - [Jeroen Ooms keynote at UseR!2021](https://www.youtube.com/watch?v=8cv2qsZ_xZw&list=PL4IzsxWztPdmHoJwIVa4um44w2GMjctmP&index=8)
  - Builds package binaries, no need to install extra packages
  - Example: LCBC's [ggseg](https://ggseg.r-universe.dev/ui#builds) universe
  - Subscribe to the [rOpenSci newsletter](https://ropensci.org/news/)
]

- locally
  - building a package creates a `tar.gz` file that can be used to install.
  
---
# Resources

- [usethis website](https://usethis.r-lib.org/)
- [Advanced R](https://adv-r.hadley.nz/) - Hadley Wickham
- [R packages](https://r-pkgs.org/) - Jenny Bryan & Hadley Wickham
- [R package primer](https://kbroman.org/pkg_primer/) - Karl Broman
- [My notes on building tidy tools from RStudio::2019](https://drmowinckels.io/blog/2019-02-02-building-tidy-tools-a-quick-recap-of-rstudio-conf-2019-workshop/)
- [My talk at the UiO IT conference demoing package creation](https://www.uio.no/for-ansatte/enhetssider/arrangement/it-konferansen/2021/) (look for the 13:00 talk on the 7th of Sept.)

---

---
background-image: url("https://drmowinckels.io/about/profile.png")
background-position: right bottom
background-size: auto 100%
class: middle, dark