Data-projects-with-R-and-GitHub

This aggregated cheatsheet of our tutorial package Data-projects-with-R-and-GitHub summarizes the individual notes from the provided tutorials and additional information from recommended sources.

Git-ing started

• git enables collaborative project development (UZH tutorial)
  • keeps track of all changes
  • maintains a local copy of a project on your computer (offline)
  • can synchronize this copy with a remote copy (e.g. online on GitHub)
    • it is able to merge multiple (text) file changes in most cases
• always pull the recent version before starting your daily work
• when you want to lock and upload your changes you have to
  1. commit the changed files to your local copy along with a useful comment
     • you can do multiple commits with independent comments (but all changes still offline)
  2. push all “committed” changed to the remote copy (e.g. GitHub)
     • Note: this might fail if the remote version of your project already has newer changes!
     • Recommended: update your project’s status via pull before doing a push!
• at some point you will get a merge conflict:
  • i.e. git finds a newer remote version of a file changed by you with changes at the same spots
  • thus, git does not know what is the correct change
  • respective lines are highlighted with arrow blocks <<<<, >>>>
    • find them
    • decide what lines to keep (and remove the others)
    • remove the <<<<, >>>>, .. parts
    • commit and push your manually merged files!
• a branch is a “copy” of your project with an independent tracking of changes
  • they are used to isolate code developments until they are stable/working
  • a pull request aims at merging your changes back into the project

Cheatsheet “coding-basics”

• important data structures
  • vector = all elements are same data type
  • list = most general, can contain anything, any size, …
  • data.frame/tibble = list of vectors (of same lengths)
    • ==> if tidy, resp. column values (in one row) are associated with one observation
• subsetting can be done using
  • [] = reduces the current object to the selected part(s) (same data container)
  • [[]] = provides a single element (element-specific data container)
  • $ = shortcut for [[]] with name
• programming of workflows often requires
  • iteration of code blocks (for or while loops)
    • for( VARIABLE in DATA ) {...}
    • each element of DATA is one by one stored in VARIABLE before running {...}
    • e.g. for( d in list( x="haha", y=1:3)) { print(length(d)) }
  • conditional code execution (if/else statements)
    • if ( CONDITION ) {..T..} else {..F..}
    • else {..B..} is optional
    • CONDITION must evaluate to a single TRUE|FALSE: triggers execution of respective {..T|F..} block
      • e.g. if ( 1:4 == 2 ) {} not working since check returns four logic values!
    • e.g. if ( version$os == "mingw32" ) { print("MS Windows user?") }
  • generalization of code blocks (function definition and usage)
    • myFunction <- function (ARGUMENTS) {...}
    • ARGUMENTS are optional (first should be the data to work on)
    • run via myFunction( .. ) using appropriate values for ARGUMENTS
    • return via return() in {...} (default: last “printed” value is returned)
    • to return multiple values, aggregate them in list() or vector c()
• redundant code w.r.t. values of a variable ==> for loop generalization
• redundant code w.r.t. variables or function calls ==> function generalization
• iteration tricks
  • best iterate directly on the elements, if you don’t need the element names or index!
  • use seq_along() to generate the list of valid indices of a vector or list
  • use names() to access the vector of element names of an object
• when a function calls itself it is called a recursive call or recursion
• function names (without brackets) can be stored in and used from variables/arguments to call the functions

How to develop a function

• first think about INPUT and OUTPUT of your function, i.e. what do you need and what do you want to create
• INPUT will be all function arguments, i.e. the needed variables of your code
• create example variables with example content
• write a piece of code that operates on your example variables doing what you want, i.e. creating your OUTPUT
• wrap your working code in a function(){} construct/body and add a return() statement to provide your OUTPUT
• test your function with your example content as argument values