Debugging tips + tools

Lecture 16

Dr. Benjamin Soltoff

Cornell University
INFO 5001 - Fall 2023

2023-10-23

Announcements

Announcements

• Continue working on group projects
• Exam begins Thursday at 8am

Debugging

Source: The Computer History Museum

Bugs

An error, flaw, failure or fault in a computer program or system that causes it to produce an incorrect or unexpected result, or to behave in unintended ways.

Artwork by @allison_horst

Approach to debugging

1. Realize that you have a bug
2. Make it repeatable
3. Figure out where it is
4. Fix it and test it

Conditions in R

1. Realize that you have a bug
2. Make it repeatable
3. Figure out where it is
4. Fix it and test it

Condition system

• Signal a condition
• Handle (or ignore) a condition

Fatal errors

addition <- function(x, y){
  if(!is_numeric(c(x, y))) stop("One of your inputs is not a number.")
  
  x + y
}

addition(3, "abc")

Error in is_numeric(c(x, y)): could not find function "is_numeric"

Warnings

logit <- function(x){
  log(x / (1 - x))
}

logit(-1)

Warning in log(x/(1 - x)): NaNs produced

[1] NaN

Warnings

logit <- function(x){
 if (x < 0 | x > 1) stop('x not between 0 and 1')
 log(x / (1 - x))
}

logit(-1)

Error in logit(-1): x not between 0 and 1

Warnings

logit <- function(x){
  x <- if_else(x < 0 | x > 1, NA_real_, x)
  if (is.na(x)) warning('x not between 0 and 1')
  log(x / (1 - x))
}

logit(-1)

Warning in logit(-1): x not between 0 and 1

[1] NA

Messages

ggplot(diamonds, aes(carat, price)) +
  geom_point() +
  geom_smooth()

`geom_smooth()` using method = 'gam' and formula = 'y ~ s(x, bs = "cs")'

Suppressing messages

demo_message <- function() message("This is a message")
demo_message()

This is a message

suppressMessages(demo_message())

demo_print <- function() print("This is a message")
demo_print()

[1] "This is a message"

suppressMessages(demo_print())

[1] "This is a message"

What if there are no conditions?

An error, flaw, failure or fault in a computer program or system that causes it to produce an incorrect or unexpected result, or to behave in unintended ways.

• Bugs can occur without any conditions in R
• If the output is unintended, it is a bug

Generate a reprex

1. Realize that you have a bug
2. Make it repeatable
3. Figure out where it is
4. Fix it and test it

Artwork by @allison_horst

Reproducible examples

• reprex (noun)
• reprex
• reprex::reprex()

Why reprexes?

Easier to talk about code that:

• Actually runs
• I don’t have to run
• I can easily run

Prior attempts to reprex

• A screenshot

• Copy-and-paste the output

• Format the output

  
  ```r
  # add code here
  ```
  

Generate a reprex using reprex()

library(tidyverse)
count(diamonds, colour)

Error in `count()`:
! Must group by variables found in `.data`.
✖ Column `colour` is not found.

Reprex do’s and don’ts

• Ensure the example is fully reproducible
• Use the smallest, simplest, most built-in data possible
• Include commands on a strict “need to run” basis
• Consider including “session info”
• Use good coding style to ensure the readability of your code by other human beings
• Ensure portability of the code

Use the call stack and/or intuition to find the bug

1. Realize that you have a bug
2. Make it repeatable
3. Figure out where it is
4. Fix it and test it

The call stack

f <- function(a) g(a)
g <- function(b) h(b)
h <- function(c) i(c)
i <- function(d) {
  if (!is.numeric(d)) {
    stop("`d` must be numeric", call. = FALSE)
  }
  d + 10
}
f("a")

Error: `d` must be numeric

traceback()

5: stop("`d` must be numeric", call. = FALSE) at #3
4: i(c) at #1
3: h(b) at #1
2: g(a) at #1
1: f("a")

Slow down, simplify, and do small things

• Run code incrementally (especially valuable for piped operations)
• Strip out extraneous operations
• Simplify to a minimal reproducible example

Unexpected output

library(tidyverse)

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point() +
  labs(x = "Displacement",
       y = "Highway MPG",
       color = "Drive") +
  scale_fill_viridis_d(option = "plasma", end = 0.9) +
  theme_minimal() +
  theme(legend.position = "bottom")

Expected to use the “plasma” color palette from viridis. What went wrong?

Source: Andrew Heiss

Simplify

• Strip it down to a basic plot
• Add simplified broken part
• Figure out broken part
• Add some parts back in
• Add the rest back in

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point()

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point() +
  scale_fill_viridis_d()

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point() +
  scale_color_viridis_d()

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point() +
  scale_color_viridis_d(option = "plasma", end = 0.9)

ggplot(data = mpg, mapping = aes(x = displ, y = hwy, color = drv)) +
  geom_point() +
  labs(x = "Displacement",
       y = "Highway MPG",
       color = "Drive") +
  scale_color_viridis_d(option = "plasma", end = 0.9) +
  theme_minimal() +
  theme(legend.position = "bottom")

Fix the bug and test it

1. Realize that you have a bug
2. Make it repeatable
3. Figure out where it is
4. Fix it and test it

Dealing with bugs

• Unexpected errors
• Expected errors
  • Add conditions
  • Use safely()

Dealing with failure using safely()

• Adverb - modifies a function (verb)
• Always returns a list with two elements
  1. result
  2. error

Dealing with failure using safely()

safe_sqrt <- safely(sqrt)
str(safe_sqrt(9))

List of 2
 $ result: num 3
 $ error : NULL

str(safe_sqrt("a"))

List of 2
 $ result: NULL
 $ error :List of 2
  ..$ message: chr "non-numeric argument to mathematical function"
  ..$ call   : language .Primitive("sqrt")(x)
  ..- attr(*, "class")= chr [1:3] "simpleError" "error" "condition"

safely() and map()

x <- list("a", 4, 5)

# unsafely square root
y <- map(x, sqrt)

Error in `map()`:
ℹ In index: 1.
Caused by error:
! non-numeric argument to mathematical function

# safely log
y <- map(x, safely(sqrt))
str(y)

List of 3
 $ :List of 2
  ..$ result: NULL
  ..$ error :List of 2
  .. ..$ message: chr "non-numeric argument to mathematical function"
  .. ..$ call   : language .Primitive("sqrt")(x)
  .. ..- attr(*, "class")= chr [1:3] "simpleError" "error" "condition"
 $ :List of 2
  ..$ result: num 2
  ..$ error : NULL
 $ :List of 2
  ..$ result: num 2.24
  ..$ error : NULL

transpose()

y <- transpose(y)
str(y)

List of 2
 $ result:List of 3
  ..$ : NULL
  ..$ : num 2
  ..$ : num 2.24
 $ error :List of 3
  ..$ :List of 2
  .. ..$ message: chr "non-numeric argument to mathematical function"
  .. ..$ call   : language .Primitive("sqrt")(x)
  .. ..- attr(*, "class")= chr [1:3] "simpleError" "error" "condition"
  ..$ : NULL
  ..$ : NULL

Application exercise

ae-14

• Go to the course GitHub org and find your ae-14 (repo name will be suffixed with your GitHub name).
• Clone the repo in RStudio Workbench, open the Quarto document in the repo, and follow along and complete the exercises.
• Render, commit, and push your edits by the AE deadline – end of tomorrow

Recap

• Debugging is a frustrating, complicated skill that you need to learn
• Use conditions or careful observations to determine if you have a bug
• Generate a reprex in order to make the bug repeatable
• Use the call stack and/or intuition to track down the bug
• Fix the bug or use condition handling to deal with it

INFO 3312/5312

Course site (from last spring)