The workflowr R package: a framework for reproducible and collaborative data science

Slide 1

Slide 1 text

The workflowr R package: a framework for reproducible and collaborative data science John Blischak (@jdblischak) 2018-07-11 useR! 2018 Brisbane, Australia github.com/jdblischak/workflowr

Slide 2

Slide 2 text

My computational challenges Organizing files Tracking intermediate results Sharing results John Blischak - github.com/jdblischak/workflowr

Slide 3

Slide 3 text

John Blischak - github.com/jdblischak/workflowr

Slide 4

Slide 4 text

Literate programming John Blischak - github.com/jdblischak/workflowr Source code Results file.Rmd file.html yihui.name/knitr rmarkdown.rstudio.com

Slide 5

Slide 5 text

R Markdown websites John Blischak - github.com/jdblischak/workflowr rmarkdown.rstudio.com

Slide 6

Slide 6 text

Version control John Blischak - github.com/jdblischak/workflowr version: 2rko6xn message: Start new… version: d1zyskv message: Update parameters… version: z6o3b97 message: Label axes… git-scm.com github.com/ropensci/git2r

Slide 7

Slide 7 text

Version control terminology repository – the tracked files and their revision history commit – a snapshot of the current state of the files John Blischak - github.com/jdblischak/workflowr

Slide 8

Slide 8 text

Web hosting GitHub Pages – hosts one website per code repository John Blischak - github.com/jdblischak/workflowr pages.github.com

Slide 9

Slide 9 text

workflowr Organized Reproducible Shareable John Blischak - github.com/jdblischak/workflowr Version-controlled websites

Slide 10

Slide 10 text

Organized John Blischak - github.com/jdblischak/workflowr

Slide 11

Slide 11 text

Start a new project > wflow_start("myproject") 1. Creates directory with template files 2. Changes working directory 3. Initiates Git repository and commits files Also available as RStudio Project Template John Blischak - github.com/jdblischak/workflowr

Slide 12

Slide 12 text

Organized directory structure John Blischak - github.com/jdblischak/workflowr R Markdown files HTML files Website options

Slide 13

Slide 13 text

Reproducible John Blischak - github.com/jdblischak/workflowr

Slide 14

Slide 14 text

Run code in clean environment John Blischak - github.com/jdblischak/workflowr > wflow_build(c("f1.Rmd", "f2.Rmd")) f1.Rmd f2.Rmd github.com/r-lib/callr

Slide 15

Slide 15 text

Tracking intermediate results > wflow_publish("analysis/file.Rmd") Performs 3-steps: 1. Commits analysis/file.Rmd 2. Builds analysis/file.Rmd 3. Commits docs/file.html and figure files John Blischak - github.com/jdblischak/workflowr

Slide 16

Slide 16 text

Combining rmarkdown and Git John Blischak - github.com/jdblischak/workflowr Source code Results 1ong9jt ln412fy Source code Results wr1q7bk 3tg6lse

Slide 17

Slide 17 text

View past results John Blischak - github.com/jdblischak/workflowr

Slide 18

Slide 18 text

Other reproducibility features output: workflowr::wflow_html Records the session information at the end Sets a seed prior to running code John Blischak - github.com/jdblischak/workflowr

Slide 19

Slide 19 text

Reproducibility report John Blischak - github.com/jdblischak/workflowr

Slide 20

Slide 20 text

Shareable John Blischak - github.com/jdblischak/workflowr

Slide 21

Slide 21 text

Slide 22

Slide 22 text

Installation 1. Install R ◦ (Recommended) Install RStudio ◦ (Optional) Install pandoc ◦ (Optional) Install Git 2. Install workflowr from CRAN ◦ install.packages("workflowr") 3. Create an account on GitHub Documentation: https://jdblischak.github.io/workflowr John Blischak - github.com/jdblischak/workflowr

Slide 23

Slide 23 text

In summary, using workflowr… Enables you to start working reproducibly immediately Allows you to focus on your analysis Shares your results online John Blischak - github.com/jdblischak/workflowr

Slide 24

Slide 24 text

Acknowledgements Co-authors: Peter Carbonetto, Matthew Stephens Early adopters for testing and feedback Authors and contributors to knitr, rmarkdown, git2r, callr John Blischak - github.com/jdblischak/workflowr

Slide 25

Slide 25 text

workflowr Organized Reproducible Shareable John Blischak - github.com/jdblischak/workflowr Version-controlled websites