Basic Debugging

• Copy and paste the example broken script into R

Basic Manual Debugging Strategy

• How do we figure out what’s wrong with a program?
  • Be a scientist.
    • Hypothesize about what is wrong.
    • Make one change that is expected to fix error.
    • Check if change worked/fixed error.
  • Do not change something without a reason.
  • Developing hypotheses (observe)
    • See where the code failed.
    • Read the error message.
    • Observe what the code is doing.
      • Look at the current state of the environment (snapshot of what’s going on)
      • Talk through the code.
      • Rubber duck programming
  • Run the code line by line checking each step.
  • Have a LLM look for problems

Example

• Create a file with the following code in it
• Run it
• Work through debugging the code

library(dplyr)
library(readr)

surveys <- read_csv('surveys.csv')
species <- read_csv('species.csv')

do_counts_by_year <- survey |>
  filter(species = "DO") |>
  group_by(year)
  summarize(count = n())

ggplot(do_count_by_year, aes(x = year, y = count)) +
  geom_point() +
  geom_line()
  labs(x = "Year", y = "Count")

Debugged version of example

library(dplyr)
library(ggplot2)
library(readr)

surveys <- read_csv('surveys.csv')

do_counts_by_year <- surveys |>
  filter(species_id == "DO") |>
  group_by(year) |>
  summarize(count = n())

ggplot(do_counts_by_year, aes(x = year, y = count)) +
  geom_point() +
  geom_line() +
  labs(x = "Year", y = "Count") 

• Now copy the code into a LLM and see what it finds and what it gets wrong