Skip to content

Webinar_notes.MD

Latest commit

 

History

History
84 lines (68 loc) · 4.89 KB

Webinar_notes.MD

File metadata and controls

84 lines (68 loc) · 4.89 KB

Git(hub) webinar OSCR – 29-04-2020

Ana Martinovici (assistant professor of marketing  attention, decision making, Bayesian statistics) Github.com/anamartinovici/githubfun

News

  • R-Ladies Netherlands book club: rladiesnl.github.io/book_club
  • How to use R and the tidyverse to clean, plot and analyze data: github.com/aschetti/ERIM2020_intro_tidyverse

Why we should use version control

Best practice: use version control from the start of the project and document your code. It will be shareable more easily.

Why version control? Some research goals:

  • Collaborate with co-authors: you can work on the same project (repository) with more people
  • Contribute to ongoing projects: this is possible even when you are not an official collaborator on the github repository
  • Share your work with the world: you can make your repos public or private and also add a (software) license to the repository
  • Reproducible results: if you can take the original data and original computer code used to analyze the data and reproduce all of the numerical findings from the study. Spectrum:

Goal: reproduce all results and nothing but the results included in the current version of the paper and previous analyses as fast and easy as possible.

  • Best case scenario: raw data > cleaning/processing > processed data > analysis > results
  • Most likely scenario: raw data v4 > cleaning/processing > processed data v10 > analysis > results v50

Things that can go wrong without version control:

  • Laptop/external HD explodes
  • Overwriting files, but also not overwriting files
  • Forgetting which of the final files is really final
  • Change file X, but forget to update all the other files/results that depend on it
  • Software changes

Version control:

  • Time travel
  • Safe place for your data and code
  • Notes to your future self

Git: formal version control system. Developed by Linus Torvalds (used to manage the course code for Linux). Tracks content: source code, data analysis projects, websites, presentations, manuscripts. Github: home for git repositories, interface for exploring public git repositories. A safe place to keep code and data. Additional benefits: issues (notes to yourself or collaborators), projects, wiki, insights.

Why use GitHub?

  • Facilitates exploring code, tracking issues, learning from others
  • Lowers the barriers to collaboration:
  • Email: “there’s a type in your code in file X line X” VS
  • Pull request: “here’s a correction to your code”
  • Free for researchers and students

How to use version control

General idea:

Vocabulary

  • Repository (repo): folder containing all tracked files as well as the version control history
  • Commit: snapshot of changes made to the staged file(s), to save a snapshot of changes made to the staged file(s)
  • Stage: staging area holds the files to be included in the next commit; to mark a file to be included in the next commit
  • Track: a tracked file is one that is recognized by the git repository
  • Branch: a parallel version of the files in a repository
  • Local: version of your repo on your pc
  • Remote: version of your repo stored on a remote server, e.g. Github
  • Clone: local copy of a remote repo on your PC
  • Fork: copy of another user’s repo on GitHub, to copy a repo to your own account. Changes made to the original repo are not changed in your forked repo (unless you pull the changes of that repo again)

See slides...

Link: Setting up Git with RStudio In RStudio: Tools > Global options > Git/SVN: make sure that Enable version control interface is checked, that R knows where to find git.

Create your first repo: online and offline

  • Internet: Go to your github account > Create a new repository
  • Give it a name (no spaces)
  • Always initialize with a readme
  • This creates an online version, to make changes, you need to clone the repo to your PC
  • Local:
  • On github, click Clone or download and copy the hyperlink
  • In RStudio, go to New project  Version control  Git  paste repository URL and choose a location to save the local copy
  • In the RStudio project, there should be a Git tab (topright panel).

Working in a local repo

  • Create files like you are used to in the folder on your local PC. Using R is not necessary, but you can just use RStudio to make commits and push changes
  • Files that are changed or newly created will appear in RStudio in the Git tab
  • In this Git tab, click the file/change that you wish to save: you will now commit the changes (“create a shipping label”)
  • Commit > in the review changes window, write a short commit message (what did you change exactly/why?) > In present tense 😊
  • Click Commit
  • Click Push
  • After pushing, the changes are also visible in your online version on GitHub