Welcome to the Software Carpentry Bootcamp at Dalhousie University!

Please make sure that you write down your name in the blank space on the upper right corner - this will help is identify the different people. 

Here we will:
1. Communicate among ourselves using the chat area (lower right corner);
2. Instructors will paste the code they are typing on the screen;
3. Share useful links;

This space is for everybody to use, feel free to add related content, fix typos and so forth. We recommend you to save a personal copy of the entire session as a text file for your reference at the end of each day.

You can add text over here...


https://github.com/dbarneche/2014-07-14-/raw/gh-pages/data/lessons.zip

Phys.ocean users: anchor:/var/tmp/lessons.zip

Dalhousie
* If there's an operation that's going to happen repeatedly, consider making it into a function.
* Give functions meaningful names.  Give EVERYTHING meaningful names.
* Indentation is important for the human reader.  (And a few languages, like Python.)
* A function should do one small, simple task.
* Compose functions from simpler functions.

EXERCISE:  Write a function called 'variance'; see formula on whiteboard. When done, compare your function's output with the builtin function 'var':
   var(x) == variance(x)

* "Premature optimization is the root of all evil" (http://c2.com/cgi/wiki?PrematureOptimization)



add.trend.line  <-  function(data, response, predictor, ...) {
  fit  <-  lm(data[[response]] ~ log10(data[[predictor]]))
  abline(fit, ...)
}

plot(lifeExp ~ gdpPercap, 
     data=data.1982, 
     log="x", cex=cex, las=1, 
     pch=21, col='black', 
     bg=continent.colour)

add.trend.line(data.1982, 
               "lifeExp", 
               "gdpPercap", 
               lty="dashed",
               col='dodgerblue2',
               lwd=2)
               
               
add.continent.trend.line  <-  function(data, continent) {
  add.trend.line(data[data$continent==continent,], 
                 "lifeExp", 
                 "gdpPercap", 
                 lty="dashed",
                 col=coloured[continent],
                 lwd=2)  
}


#new implementation of add.trend.line
add.trend.line  <-  function(data, response, predictor, ...) {
  lx   <-  log10(data[[predictor]])
  fit  <-  lm(data[[response]] ~ lx)
  xr   <-  range(lx)
  lines(10^xr, predict(fit, list(lx=xr)), ...)
}

get.n.countries <- function(x) {
  n <- unique(x$country)
  length(n)
}
get.n.countries(data)

data.new <- data[data$continent == "Asia", ]
get.n.countries(data.new)
data.new <- data[data$continent == "Africa", ]
get.n.countries(data.new)

daply(data, .(continent), get.n.countries)
daply(data, .(continent),
      function(x) length(unique(x$country)))

get.total.pop <- function(x) {
  
}

add.trend.line <- function(x, y, d, ...) {
  fit <- lm(d[[y]] ~ log10(d[[x]]))
  abline(fit, ...)
}

colour.by.category <- function(x, table) {
  unname(table[x])
}

rescale <- function(x, r.out) {
  p <- (x - min(x)) / (max(x) - min(x))
  r.out[1] + p * (r.out[2] - r.out[1])
}
data.1982 <- data[data$year == 1982,]
col.table <- c(Asia="tomato", Europe="chocolate4", Africa="dodgerblue2", Americas="darkgoldenrod1", Oceania="green4")

col <- colour.by.category(data.1982$continent,
                          col.table)
cex <- rescale(sqrt(data.1982$pop), c(0.2, 10))

plot(lifeExp ~ gdpPercap, data = data.1982,
     log="x", cex = cex, col = col, pch = 21)
d_ply(data.1982, .(continent),
      function(x) {
         add.trend.line("gdpPercap", 
                        "lifeExp",
                        x,
                        col = col.table[x$continent])
         })

Day 2

Testing:

variance <- function(x) {
  n <- length(x)  # number of observations
  xbar <- mean(x) # mean of x
  sum((x - xbar)^2) / (n-1) # var(x)
}

skewness <- function(x) {
  n <- length(x)
  xbar <- mean(x)
  skew <- sum((x - xbar)^3) / (n-2)
  skew <- skew / var(x)^(3/2)
  skew
}

# Proposed tests:
# 1. Skewness of a really big rnorm should be close to zero
# 2. Random log-normal values should have positive skewness

EXERCISE:  With the 'rescale' function,
1) write a unit test that checks that it gives the right range
2) does it handle a bad range input?

More motivation to test.  Carefully.
http://boscoh.com/protein/a-sign-a-flipped-structure-and-a-scientific-flameout-of-epic-proportions.html

Some links for more on testthat:

Hadley Wickham wrote an intro paper for the R Journal on testthat, which goes through the basics: http://journal.r-project.org/archive/2011-1/RJournal_2011-1_Wickham.pdf [pdf]
Hadley has also been writing a book, Advanced R, that is available as a website and which has a section on Testing using testthat: http://adv-r.had.co.nz/Testing.html  The book should be published in hardcopy form by CRC Press later in the year: http://www.crcpress.com/product/isbn/9781466586963

SHELL

cd   --- change directory
cd ..    ---- go to parent directory
pwd   --- print working directory
ls   --- list contents of directory
ls -l   --- "long" list, show details
ls --help

Exercise:
List all files/folders with the digit '3' in their name

Windows users, please run:

echo "export TERM=msys" >> ~/.bashrc



#VERY IMPORTANT
Hi everyone, whenever you can, can I please get you to re-download the material from the link indicated at the top of this page?
13:06Diego Barneche: I have updated the content for the reproducibility session, it would be great if you could have that running.
13:06Diego Barneche: Just make sure not to overwrite the work you've done so far. You only need to overwrite the folder 90-reproducibility

https://github.com/dbarneche/2014-07-14-Dalhousie/raw/gh-pages/data/lessons.zip



Additional readings on R material:
http://cran.r-project.org/doc/manuals/R-intro.html#Writing-your-own-R-intro.html#Writing-your-own-functions
http://nicercode.github.io/intro/writing-functions.html
http://adv-r.had.co.nz/

Shameless plug: Remi wrote an R tutorial for ocean science, it has some things you may find useful (like mapping):
https://github.com/remi-daigle/R_tutorial_ocean


VERSION CONTROL

echo "export TERM=msys" >> ~/.bashrc

Motivation, and chuckles:  http://www.phdcomics.com/comics/archive.php?comicid=1531

git package for MacOS Snow Leopard 10.6.8:
https://code.google.com/p/git-osx-installer/downloads/detail?name=git-1.8.4.2-intel-universal-snow-leopard.dmg

git config --global user.name "User Name"
git config --global user.email "user@foobar.com"

Windows only:
git config --global core.autocrlf "true"

Mac OS X / Linux
git config --global core.autocrlf "input"

If you have Nano:
git config --global core.editor "nano --tempfile"
If 'notepad' is the editor that works for you (Windows), try:
   git config --global core.editor "notepad"

git config --global color.ui "auto"

git init
git add file
git commit -m "message"
git status
git log

git branch
git branch new_branch
git checkout another_branch
git branch -d branch_to_be_deleted

To revert to a previous git commit id

git reset --hard commit-hash for example:

git checkout -b master
git reset --hard bc56e0e

will reset the master branch to the state as of the commit bc56e0e

REPRODUCIBLE RESEARCH

Goals of reproducible research

to.pdf <- function(expr, filename, ..., verbose = TRUE) {
    if (verbose)
        cat(sprintf("Creating %s\n", filename))
    pdf(filename, ...)
    on.exit(dev.off())
    eval.parent(substitute(expr))
}

dir.create("output/figures")

to.pdf(myplot(data.1982, "gdpPercap", "lifeExp", main = "1982", lty = 2),
           "output/figures/figure1.pdf", height = 7, width = 7)

for (i in unique(data$year)) {
    subdata <- data[data$year == i,  ]
    to.pdf(myplot(subdata, "gdpPercap", "lifeExp", main = "1982", lty = 2),
               paste0("output/figures/figure1_", i, ".pdf", height = 7, width = 7)
}

R packages used:

knitr: the knitr website http://yihui.name/knitr/

Stackoverflow is a great source of help for all kinds of coding issues:

stackoverflow.com


SUM-UP

http://dbarneche.github.io/2014-07-14-Dalhousie

What we're aiming at: "Best Practices For Scientific Computing" http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001745