#Tutorial 1: Basics of R for non-programmers
##SUMMARY
- Running R / RStudio
- Working with packages
- learning R with swirl
- Installing github packages
- tips
##RUNNING R
Assuming R is installed...
First thing, we need to have a look at the different ways of running R
- from a command line
- in Rstudio
RStudio is a great way to start, especially if you are used to GUIs. But there are occasions (such as when you are working remotely) where the command line is useful.
OK, let's start RStudio. For more help, see:
[https://www.rstudio.com/resources/training/online-learning/]
##WORKING WITH PACKAGES
There are five levels of programming in R:
- your own code Programs you'd only use yourself, for analysing and presenting your own data. Before we do any programming, we should generally check if a solution exists elsewhere!
- shared code (bad practice) Programs you share locally with colleagues - proprietary stuff. It is very easy to email programs to one another, but often this can cause problems if the programs don't work for one reson or another. It is very easy to lose track of the 'correct' version of your code, and it is often poorly documented.
- packages It's not much more work to group your programs together into a package that you develop as a group. We'll show how to do this using github below. The advantages are that it is easier to document what you are doing as you go along, and you can timestamp versions of the code so it is easy to keep track of how things work at any point in the development of the software.
- CRAN packages Next level ('The Comprehensive R Archive Network'). Specialist programs, maintained by the R community. Rigorously tested, very powerful, but sometimes not very well documented and quirky to use. The requirements for CRAN hosting can be quite challenging to meet, but that's why it's the 'gold standard'!
- R 'Core' This is the R programming language, operators functions etc. We'll have a look at how to learn this stuff over the next few sections
###INSTALLING PACKAGES
Most of the time, you'll be installing packages from CRAN, since these are the most tried and tested ones. These are the easiest packages to install.
I'm going to show you two packages that you should know about:
- Swirl - Friendly R tutorials
- devtools - ways to help you oragnise your code into packages from the outset. (More of this)
##Learning R with Swirl
Packages can be about anything. 'Swirl' is a good package to start with, because it contains tutorials for learning R.
Let's install swirl using the RStudio menu:
tools->install packages...
gets you to a dialog box. Type swirl into the 'Packages' text-entry field.
Alternatively, at the command line (in the bottom left panel of RStudio) type
install.packages("swirl")
####Troubleshooting write permissions problems
If you are using RStudio on a Windows machine where you do not have admin privileges, you may encounter problems when installing packages, because R might try and put the package files in a directory that you cannot write to. If you get an error indicating problems with write permissions, have a look at the following, from [http://stackoverflow.com/questions/15170399/changing-r-default-library-path-using-libpaths-in-rprofile-site-fails-to-work] :
"The proper solution is to set environment variable R_LIBS_USER to the value of the file path to your desired library folder as opposed to getting RStudio to recognize a Rprofile.site file.
To set environment variable R_LIBS_USER in Windows, go to the Control Panel (System Properties -> Advanced system properties -> Environment Variables -> User Variables) to a desired value (the path to your library folder), e.g.
Variable name: R_LIBS_USER
Variable value: C:/software/Rpackages
Setting R_LIBS_USER will ensure that the library shows up first in .libPaths() regardless of starting RStudio directly or by right-clicking an file and "Open With" to start RStudio."
##Loading the package
REMEMBER! Installing a package is not the same as loading a package so that you can use it. Think of installing a package as like buying a book and putting it on your bookshelf. You can't read it while it's on the shelf - you have to hold it first. Similarly, you have to load the package before you can use it, using the library command:
library("swirl")
For more on swirl, see:
[http://swirlstats.com/students.html]
This is a set of interactive tutorials. Let's install that and go through a bit
There's also an online course starting 18th Jan 2015 - it's free - coursera are pretty good:
[https://www.coursera.org/learn/r-programming]
##Installing github packages
Packages that are not hosted by CRAN require a little more work to install. Fortunately, the 'devtools' package helps with this. Devtools is hosted by CRAN, so we can install it in the same way we installed Swirl.
devtools is a great thing for building packages. If you plan to share your software, you should begin by creating a package. It will save you time in the long run! If you've ever come back to some code, or even a complicated spreadsheet and wondered how it works, you should know what I mean. Using devtools and another package called roxygen makes it super easy to maintain your programs in a package.
To demonstrate this, we're going to load the bioarch package that this tutorial came with.
Here's the command-line for installing and loading devtools (you can use the RStudio menu if you like):
> install.packages("devtools")
> library(devtools)
Then you can use the install_github function to install bioarch:
> install_github("franticspider/bioarch",subdir = "bioarch")
Of course, you'll only need to do that once for your R installation. After that, whenever you want to use bioarch, simply enter:
> library("bioarch")
We'll talk further about this next time, but rember, if a package is written entirely in R, then install will work no matter what platform you are using. Extra steps have to be taken if the R package uses programs written in other languages (yes, R can do that) For example, the q2e package we've built uses Julie Wilson's C code, so we have to take a few more steps to make it work in Windows Here's a great blog post about writing packages (if a little out of date):
[http://hilaryparker.com/2014/04/29/writing-an-r-package-from-scratch/]
##Tips
###getting help
This is much easier in Rstudio than the command line. Hit the 'help' tab in the bottom right panel. The help files for each package is shown. If you've just installed a package, you might need to hit the refresh icon to see it.
###Listing packages
installed.packages()
will list everything.
###Removing packages
should be simple:
remove.packages("devtools")
If the package isn't there, you'll get this error:
Error in remove.packages : there is no package called ‘devtools’
sometimes the package to be removed isn't in the usual directory. If you think it's still there do this:
.libPaths()
which will show all the paths that packages can be on. You can list these directories in the remove step like this:
remove.packages("devtools","/usr/lib/R/library")