Code Review and Testing with gooseR

In gooseR: R Integration for 'Goose' AI

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = FALSE
)

library(gooseR)

Introduction

gooseR provides intelligent code review that goes beyond static analysis. Unlike traditional linters that check syntax, gooseR actually reads and understands your code, providing context-aware feedback based on what you're trying to accomplish.

The Power of goose_honk()

The goose_honk() function is your intelligent code reviewer. It offers four severity levels, each with a different personality and focus:

Severity Levels

# Gentle - Encouraging and constructive
goose_honk(severity = "gentle")
# "I notice you're using a loop here. Have you considered using lapply()? 
#  It might be more efficient! Your code structure looks good overall! 🦆"

# Moderate - Balanced and professional
goose_honk(severity = "moderate")
# "Your loop on line 15 could be replaced with vectorized operations for 
#  better performance. Also, consider adding error handling for the mean() 
#  function in case of NA values."

# Harsh - Direct and critical
goose_honk(severity = "harsh")
# "That loop is killing your performance. Use vectorization. No error 
#  handling? That's asking for production failures. Fix the variable 
#  naming - 'x' and 'df' tell me nothing."

# Brutal - No holds barred
goose_honk(severity = "brutal")
# "This code is a disaster. Loops in R? Really? No error handling, 
#  meaningless variable names, and zero documentation. Did you even 
#  test this? Start over and do it right."

Real-World Example: Data Analysis Script

Let's see how goose_honk() helps improve a real analysis:

# A typical analysis script with issues
analyze_sales <- function(sales_data) {
  # Calculate totals
  total = 0
  for(i in 1:nrow(sales_data)) {
    total = total + sales_data$amount[i]
  }

  # Get average
  avg = mean(sales_data$amount)

  # Find best month
  best = sales_data[sales_data$amount == max(sales_data$amount),]

  # Make plot
  plot(sales_data$amount)

  return(list(total, avg, best))
}

# Get gentle feedback first
goose_honk(severity = "gentle")

Gentle Review Output:

🦆 Hey there! Your analyze_sales function is a great start! Here are some friendly suggestions:

✨ Opportunities for improvement:
• The loop for calculating total could be replaced with sum(sales_data$amount) - much faster!
• Consider checking for NA values in your mean calculation: mean(sales_data$amount, na.rm = TRUE)
• Your plot could use some labels to make it clearer

💡 Quick wins:
• Named list elements would make your return value clearer
• Adding input validation would make your function more robust

Keep up the good work! Your logic is sound! 🌟

# Now let's get more critical feedback
goose_honk(severity = "harsh")

Harsh Review Output:

Your analyze_sales function has serious issues:

PERFORMANCE PROBLEMS:
- Line 3-6: That loop is inexcusable. Use sum(sales_data$amount). You're writing R like it's C.
- This will be 100x slower on large datasets.

BUGS WAITING TO HAPPEN:
- No NA handling in mean() - this WILL break in production
- max() without na.rm = TRUE is a ticking time bomb
- What if multiple months have the same max value? Your subsetting is naive.

POOR PRACTICES:
- Unnamed return list? How is anyone supposed to use this?
- No input validation - what if sales_data is NULL? Empty? Missing columns?
- Base R plotting in 2024? At least use ggplot2.

Fix these issues before this sees production.

Improved Version After Review

Based on the feedback, here's an improved version:

analyze_sales <- function(sales_data) {
  # Input validation
  if (is.null(sales_data) || nrow(sales_data) == 0) {
    stop("sales_data cannot be NULL or empty")
  }
  if (!"amount" %in% names(sales_data)) {
    stop("sales_data must contain 'amount' column")
  }

  # Calculate metrics with NA handling
  total_sales <- sum(sales_data$amount, na.rm = TRUE)
  avg_sales <- mean(sales_data$amount, na.rm = TRUE)

  # Find best months (handle ties)
  max_amount <- max(sales_data$amount, na.rm = TRUE)
  best_months <- sales_data[sales_data$amount == max_amount & 
                           !is.na(sales_data$amount), ]

  # Create informative visualization
  library(ggplot2)
  p <- ggplot(sales_data, aes(x = seq_along(amount), y = amount)) +
    geom_line() +
    geom_point() +
    theme_brand("block") +
    labs(title = "Sales Trend", x = "Period", y = "Sales Amount")

  print(p)

  # Return named list
  return(list(
    total = total_sales,
    average = avg_sales,
    best_months = best_months,
    plot = p
  ))
}

# Check our improvements
goose_honk(severity = "moderate")

Context-Aware Analysis

goose_honk() understands different types of R code:

Data Manipulation

# It recognizes dplyr chains
result <- data %>%
  filter(x > 10) %>%
  group_by(category) %>%
  summarise(mean = mean(value))

goose_honk()
# "Good use of dplyr! Consider adding .groups = 'drop' to summarise() 
#  to avoid the grouped data frame warning."

Statistical Models

# It understands modeling
model <- lm(mpg ~ wt + cyl, data = mtcars)

goose_honk()
# "Linear model looks good. Have you checked assumptions? 
#  Consider plot(model) for diagnostics. Also, you might want 
#  to check for multicollinearity between wt and cyl."

Visualization

# It recognizes ggplot2
p <- ggplot(data, aes(x, y)) + geom_point()

goose_honk()
# "Basic scatter plot. Consider adding labels with labs(), 
#  applying a theme, and perhaps adding a trend line with 
#  geom_smooth() if appropriate."

Generating Tests

gooseR can generate test suites for your functions:

# Your function
calculate_bmi <- function(weight_kg, height_m) {
  if (height_m <= 0) stop("Height must be positive")
  if (weight_kg <= 0) stop("Weight must be positive")

  bmi <- weight_kg / (height_m ^ 2)

  category <- if (bmi < 18.5) "Underweight"
  else if (bmi < 25) "Normal"
  else if (bmi < 30) "Overweight"
  else "Obese"

  return(list(bmi = bmi, category = category))
}

# Generate tests
tests <- goose_generate_tests("calculate_bmi")
cat(tests)

Generated Tests:

test_that("calculate_bmi works correctly", {
  # Test normal case
  result <- calculate_bmi(70, 1.75)
  expect_equal(result$bmi, 22.86, tolerance = 0.01)
  expect_equal(result$category, "Normal")

  # Test edge cases
  expect_equal(calculate_bmi(50, 1.8)$category, "Underweight")
  expect_equal(calculate_bmi(85, 1.75)$category, "Overweight")
  expect_equal(calculate_bmi(100, 1.7)$category, "Obese")

  # Test error conditions
  expect_error(calculate_bmi(0, 1.75), "Weight must be positive")
  expect_error(calculate_bmi(70, 0), "Height must be positive")
  expect_error(calculate_bmi(-70, 1.75), "Weight must be positive")

  # Test boundary conditions
  result_boundary <- calculate_bmi(56.25, 1.5)  # Exactly BMI = 25
  expect_equal(result_boundary$bmi, 25)
})

Generating Documentation

Let gooseR write your roxygen2 documentation:

# Your function
clean_text <- function(text, remove_numbers = FALSE, lowercase = TRUE) {
  if (lowercase) text <- tolower(text)
  text <- gsub("[[:punct:]]", " ", text)
  if (remove_numbers) text <- gsub("[0-9]", "", text)
  text <- gsub("\\s+", " ", text)
  trimws(text)
}

# Generate documentation
docs <- goose_document("clean_text")
cat(docs)

Generated Documentation:

#' Clean and Preprocess Text Data
#'
#' Performs text cleaning operations including punctuation removal,
#' case conversion, number removal, and whitespace normalization.
#'
#' @param text Character vector. The text to be cleaned.
#' @param remove_numbers Logical. If TRUE, removes all numeric characters
#'   from the text. Default is FALSE.
#' @param lowercase Logical. If TRUE, converts all text to lowercase.
#'   Default is TRUE.
#'
#' @return Character vector of the same length as input with cleaned text.
#'   Punctuation is replaced with spaces, multiple spaces are collapsed
#'   to single spaces, and leading/trailing whitespace is removed.
#'
#' @examples
#' clean_text("Hello, World! 123")
#' # [1] "hello world 123"
#' 
#' clean_text("Hello, World! 123", remove_numbers = TRUE)
#' # [1] "hello world"
#' 
#' clean_text("HELLO WORLD", lowercase = FALSE)
#' # [1] "HELLO WORLD"
#'
#' @export

Error Explanation and Fixes

When you encounter errors, gooseR can help:

# You get an error
data <- read.csv("myfile.csv")
model <- lm(y ~ x1 + x2 + x3, data = data)
# Error: object 'y' not found

# Get help
error_help <- goose_explain_error()
cat(error_help)

Error Explanation:

The error "object 'y' not found" means the column 'y' doesn't exist in your data frame.

POSSIBLE CAUSES:
1. Column name mismatch (check names(data))
2. Column was renamed during import
3. CSV has different column names than expected

SOLUTIONS:
# Check column names
names(data)

# If column exists with different name:
model <- lm(actual_column_name ~ x1 + x2 + x3, data = data)

# Or rename the column:
names(data)[names(data) == "old_name"] <- "y"

# Defensive approach:
if (!"y" %in% names(data)) {
  stop("Column 'y' not found in data. Available columns: ", 
       paste(names(data), collapse = ", "))
}

Best Practices Workflow

Here's a complete development workflow with gooseR:

# 1. Write your function
my_function <- function(data) {
  # Initial implementation
  result <- process_data(data)
  return(result)
}

# 2. Get initial review
goose_honk(severity = "gentle")

# 3. Improve based on feedback
my_function <- function(data) {
  # Improved implementation with error handling
  if (is.null(data)) stop("Data cannot be NULL")
  result <- process_data(data)
  return(result)
}

# 4. Get stricter review
goose_honk(severity = "moderate")

# 5. Generate documentation
docs <- goose_document("my_function")

# 6. Generate tests
tests <- goose_generate_tests("my_function")

# 7. Final review
goose_honk(severity = "harsh")

# 8. Save your work
goose_save(my_function, category = "functions", tags = c("reviewed", "tested"))

Integration with RStudio/Positron

Use the addins for quick access:

1. Quick Review: Select code and use the "Review Code" addin
2. Generate Docs: Place cursor in function and use "Document Function" addin
3. Explain Error: When you hit an error, use "Explain Last Error" addin

Tips for Effective Code Review

1. Start Gentle: Begin with gentle reviews to build confidence
2. Progress Gradually: Move to harsher reviews as code improves
3. Focus on Patterns: goose_honk() identifies recurring issues
4. Learn from Feedback: Each review teaches best practices
5. Review Often: Regular reviews during development, not just at the end

Conclusion

gooseR's code review and testing features transform you into a better R programmer. The context-aware feedback, automatic test generation, and documentation creation save time while improving code quality.

For more information about gooseR's capabilities, see the other vignettes in the package documentation.

gooseR documentation built on Feb. 6, 2026, 5:07 p.m.