R Anti Slop

R Anti-Slop: Stop Writing df <- data

When to Use This

Use this for:

• ✓ Any R code leaving your machine (analysis, packages, scripts)
• ✓ AI-generated code review (catches df, result, missing ::)
• ✓ CRAN submissions (they'll reject generic code anyway)
• ✓ Team code standards

Skip for:

• Quick console experiments (though habits form fast)
• Legacy code you can't touch
• Bioconductor or other style guides that override this

Quick Example

Before (AI Slop):

# Load the library
library(dplyr)

# Read the data
df <- read.csv("data.csv")

# Filter the data
result <- df %>% filter(x > 0)

After (Anti-Slop):

customer_data <- readr::read_csv("data/customers.csv")

active_customers <- customer_data |>
  dplyr::filter(status == "active", revenue > 0)

return(active_customers)

What changed:

• ✓ Descriptive names (customer_data not df)
• ✓ Namespace qualification (dplyr::, readr::)
• ✓ Native pipe (|> not %>%)
• ✓ No obvious comments
• ✓ Explicit return

When to Use What

Core Workflow

5-Step Quality Check

1. Namespace qualification - All external functions use ::

   # Good
   dplyr::filter(data, x > 0)
   # Bad
   filter(data, x > 0)
   

2. Explicit returns - Every function has return()

   # Good
   my_function <- function(x) {
     result <- x + 1
     return(result)
   }
   # Bad
   my_function <- function(x) {
     x + 1
   }
   

3. Naming conventions - All objects use snake_case

   # Good
   customer_lifetime_value <- calculate_clv(data)
   # Bad
   df <- calculate_clv(data)
   customerLifetimeValue <- calculate_clv(data)
   

4. Documentation quality - No generic descriptions

   # Good
   #' @param deaths Data frame with `age_group` and `count` columns
   # Bad
   #' @param data The data
   

5. Code formatting - Run styler and lintr

   styler::style_file("script.R")
   lintr::lint("script.R")
   

Quick Reference Checklist

Before committing R code, verify:

• All external functions qualified with ::
• All functions have explicit return()
• All objects use snake_case
• No generic names (df, data, result, temp)
• Pipes (|>) have space before, end lines
• Long pipelines (>8 ops) broken into named steps
• Complex operations have WHY comments
• Data validated after transformations
• Seeds set before random operations
• Uncertainty reported (SE, CI) for statistical models
• No attach() calls
• No right-hand assignment (->)
• Roxygen documentation is specific
• Examples are realistic and run

Common Workflows

Workflow 1: Clean Up AI-Generated R Script

Context: AI generated an analysis script with generic patterns.

Steps:

1. Run detection script

   Rscript toolkit/scripts/detect_slop.R analysis.R --verbose
   

2. Fix high-priority issues first

   # Replace df, data, result with descriptive names
   # Before
   df <- readr::read_csv("data.csv")
   result <- df %>% filter(x > 0)
   
   # After
   customer_data <- readr::read_csv("data/customers.csv")
   active_customers <- customer_data |> dplyr::filter(status == "active")
   

3. Add namespace qualification

   # Before
   data %>% filter(x > 0) %>% summarize(mean(y))
   
   # After
   data |>
     dplyr::filter(x > 0) |>
     dplyr::summarize(mean_y = mean(y))
   

4. Add explicit returns

   # Before
   calculate_rate <- function(numerator, denominator) {
     numerator / denominator
   }
   
   # After
   calculate_rate <- function(numerator, denominator) {
     rate <- numerator / denominator
     return(rate)
   }
   

5. Break long pipes

   # Before (12 operations in one chain)
   result <- data |>
     filter(...) |> mutate(...) |> group_by(...) |>
     summarize(...) |> arrange(...) |> [7 more ops]
   
   # After
   clean_data <- data |>
     dplyr::filter(!is.na(value)) |>
     dplyr::mutate(category = categorize(value))
   
   summary_stats <- clean_data |>
     dplyr::group_by(category) |>
     dplyr::summarize(mean_val = mean(value))
   

6. Format and validate

   styler::style_file("analysis.R")
   lintr::lint("analysis.R")
   

Expected outcome: Score drops from 60+ to <20

Workflow 2: Fix Generic Package Documentation

Context: R package has generic roxygen documentation.

Steps:

1. Identify generic patterns

   # Bad
   #' Process Data
   #'
   #' @description This function processes the data.
   #' @param data The data.
   #' @return The result.
   

2. Make description specific

   # Good
   #' Calculate age-adjusted mortality rates
   #'
   #' Computes mortality rates per 100,000 population, standardized to the
   #' 2000 US Census age distribution using direct standardization.
   

3. Describe parameter structure

   # Good
   #' @param deaths Data frame with columns `age_group` and `count`.
   #' @param population Data frame with columns `age_group` and `pop_size`.
   

4. Specify return value

   # Good
   #' @return A tibble with columns:
   #'   \describe{
   #'     \item{county}{County FIPS code}
   #'     \item{rate}{Age-adjusted rate per 100,000}
   #'     \item{se}{Standard error of the rate}
   #'   }
   

5. Add realistic examples

   # Good
   #' @examples
   #' counties <- data.frame(
   #'   county = c("A", "B"),
   #'   deaths = c(150, 200),
   #'   population = c(50000, 80000)
   #' )
   #'
   #' adjust_rates(counties, rate_per = 100000)
   #' #> # A tibble: 2 x 3
   #' #>   county  rate    se
   #' #> 1 A       312.  25.4
   #' #> 2 B       258.  18.2
   

Expected outcome: Documentation that teaches, not restates

Workflow 3: Prepare Package for CRAN

Context: Final checks before CRAN submission.

Steps:

1. Run all quality checks

   # Standard checks
   devtools::check()
   
   # Anti-slop checks
   lapply(list.files("R", full.names = TRUE), function(f) {
     system(paste("Rscript toolkit/scripts/detect_slop.R", f))
   })
   

2. Fix documentation

   • Check all @param descriptions are specific
   • Verify @examples run and are realistic
   • Ensure @return describes structure

3. Validate code quality

   # Format all files
   styler::style_dir("R/")
   
   # Check lints
   lintr::lint_package()
   

4. Check CRAN-specific requirements

   • Validate URLs in DESCRIPTION and documentation
   • Check examples run in < 5 seconds
   • Verify package structure meets CRAN standards

Expected outcome: Clean R CMD check with no slop patterns

Mandatory Rules Summary

1. Namespace Qualification

ALWAYS use :: for external packages

Exceptions (don't need ::):

• Base R: mean(), sum(), log(), etc.
• stats: lm(), glm(), t.test(), etc.
• utils: head(), tail(), str(), etc.

2. Explicit Returns

ALWAYS use return() - never implicit

3. Naming: snake_case

All objects use snake_case

• Variables: customer_data not customerData or df
• Functions: calculate_rate not calculateRate
• Arguments: input_data not inputData

4. Native Pipe

Prefer |> over %>% (unless R < 4.1)

5. No Generic Names

Never use: df, data, result, temp, x, n (except standard math notation)

Tidyverse Philosophy

Follow Tidyverse Style Guide as primary reference:

1. Design for humans - Code should be readable and intuitive
2. Reuse existing data structures - Work with tibbles and data frames
3. Compose simple functions with pipes - Build complexity through composition
4. Embrace functional programming - Functions are first-class objects

See reference/tidyverse.md for complete tidyverse conventions.

Resources & Advanced Topics

Reference Files

• reference/naming.md - Complete naming conventions and forbidden patterns
• reference/tidyverse.md - Pipe conventions, formatting, ggplot2 standards
• reference/documentation.md - Roxygen2, vignettes, README quality
• reference/statistical-rigor.md - Validation, uncertainty, reproducibility
• reference/forbidden-patterns.md - Complete antipattern catalog

Related Skills

• text/anti-slop - For cleaning prose in documentation
• quarto/anti-slop - For cleaning vignettes and documentation

Tools

• styler::style_file() - Auto-format code
• lintr::lint() - Check code quality
• Rscript toolkit/scripts/detect_slop.R - Detect AI patterns

Integration with Posit Skills

This skill focuses on code quality and avoiding generic patterns.

Use together with Posit skills for complete coverage: