r-package-spellchecker

name	r-package-spellchecker
description	Spellcheck an R package with cspell, fix genuine typos in source files, and add legitimate domain words to .cspell/project-words.txt. Use when the user asks to spellcheck the package, run cspell, fix typos, or clean up the project dictionary.

R package spellchecker

Spellcheck the package with cspell, fix genuine typos, and add legitimate words to the shared project dictionary.

The cspell configuration lives in cspell.json at the package root. It checks markdown, plaintext, and .qmd files, plus comments (including roxygen) in R files. The shared dictionary is .cspell/project-words.txt.

Workflow

1. Run cspell from the package root (the config is auto-detected):

   # Issues with file/line locations for context
   npx --yes cspell --no-progress --relative .
   
   # Deduplicated list of unknown words
   npx --yes cspell --no-progress --words-only --unique . | sort -fu
   

2. Triage each unknown word using its source context (read the surrounding line in the reporting file when unsure):

   • Genuine typo → fix it in the source file. Never add a typo to the dictionary.
   • Legitimate word → append it to .cspell/project-words.txt. Examples: package, class, and function names (bbotk, Classif, clbk), math and optimization jargon (codomain, BOBYQA), author surnames from the bibliography (Bischl, Birattari), and established abbreviations (evals).
   • When a word is ambiguous, prefer fixing over adding, and mention the judgment call in the final report.

3. Apply the fixes, respecting these rules:

   • The language is en-US. British spellings are typos to fix, not words to add.
   • Never edit generated files. Fix typos in README.Rmd and regenerate with Rscript -e "devtools::build_readme()", never edit README.md directly. man/ and *.Rd are already ignored by the config.
   • After fixing typos in roxygen comments, run Rscript -e "devtools::document()".
   • Keep .cspell/project-words.txt one word per line, sorted alphabetically case-insensitively, below the existing header comments.

4. Verify: re-run cspell until it reports no unknown words and exits 0.

5. Report: summarize the typos fixed (file: old → new) and the words added to the dictionary.