01/11/2021

What is debugging?

The process of stepping through your code to find the source of an unexpected problem. This typically involves:

  1. Running the code.
  2. Stopping the code where something suspicious is taking place.
  3. Looking at the code step-by-step from this point on to either change the values of some variables, or modify the code itself.

How do we figure out what’s wrong with a program?

  • Read the error message.

    • Hypothesize about what is wrong.
    • Make one change that is expected to fix error.
    • Check if change worked/fixed error.

  • Talk through the code.

    • Observe what the code is doing.
    • Rubber duck programming

Do not change something without a reason.

Example: Debugging a for loop

Estimating the probability distribution of the numbers of suits in a random draw of five poker cards.

cards <- rep(c("diamonds","clubs","spades","hearts"), 13)

for(i in 1:10000){

  # random draw of 5 cards
  s <- sample(cards, 5)
  
  # number of suits in this draw
  n <- length(s)
  
  # increment observation of n
  obs[n] <- obs[n] + 1
}

barplot(obs/sum(obs),
        main = "Number of suits in a random draw of 5 poker cards",
        xlab = "Number of suits",
        ylab = "Estimated probability",
        names.arg = 1:4)

What happen if we source this code?

Example: Debugging a for loop

Estimating the probability distribution of the numbers of suits in a random draw of five poker cards.

cards <- rep(c("diamonds","clubs","spades","hearts"), 13)

for(i in 1:10000){

  # random draw of 5 cards
  s <- sample(cards, 5)
  
  # number of suits in this draw
  n <- length(s)
  
  # increment observation of n
  obs[n] <- obs[n] + 1
}
## Error in eval(expr, envir, enclos): object 'obs' not found
barplot(obs/sum(obs),
        main = "Number of suits in a random draw of 5 poker cards",
        xlab = "Number of suits",
        ylab = "Estimated probability",
        names.arg = 1:4)
## Error in barplot(obs/sum(obs), main = "Number of suits in a random draw of 5 poker cards", : object 'obs' not found

Example: Debugging a for loop

Manually

How can we trace mistakes…?

Here three basic debugging techniques:

  1. Reduce number of iterations

  2. Inspect variables during the loop with print().

  3. If the for-loop makes calls to R’s random generator, set seed. (e.g., sample() or rnorm(). So fix a seed before the loop so that results are reproducible).

Example: Debugging a for loop

With R built-in functions

Debugging a for loop with debugonce()

function.for <- function(nsim=10000) {

  # This function estimates the probability distribution of the numbers of suits
  # in a random draw of five poker cards.
  cards <- rep(c("diamonds","clubs","spades","hearts"), 13)

  for(i in 1:nsim){

    # random draw of 5 cards
    s <- sample(cards, 5)

    # number of suits in this draw
    n <- length(s)

    # increment observation of n
    obs[n] <- obs[n] + 1
  }

  barplot(obs/sum(obs),
        main = "Number of suits in a random draw of 5 poker cards",
        xlab = "Number of suits",
        ylab = "Estimated probability",
        names.arg = 1:4)

  return(obs)
}

Base R debug functions

debugonce()

debug(), undebug()

browser()

traceback()

options(error = browser), options(error = NULL)

RStudio IDE debugging tools

  • Editor breakpoint

  • Browser breakpoint

  • Debug mode

    • source pane highlighting
    • context-specific Environment pane, including a Traceback panel
    • browser prompt into Console pane

  • “Show traceback” shortcut

  • “Rerun with debug” shortcut

  • Debug menu

Further reading