I have a shell alias that counts lines of source code with find and wc. In a monorepo with several apps under apps/, the total came out about 100,000 lines too high. The end of the output showed why. It counted files in each app’s node_modules, in dist/ build output, and in .wrangler/ temp bundles. It also skipped .ts and .tsx files, which is most of the code in that repo.
The alias excluded node_modules like this:
-not -path "./node_modules/*"
-path matches the whole path from where find started, so ./node_modules/* only matches the node_modules directory at the top. ./apps/web/node_modules/... doesn’t match. A leading * matches at any depth:
alias loc='find . -not -path "*/node_modules/*" -not -path "*/dist/*" -not -path "*/.wrangler/*" \( -name "*.rb" -o -name "*.js" -o -name "*.ts" -o -name "*.tsx" \) | xargs wc -l'
-not -path still walks into those directories and throws the results away. On a big tree, -prune is faster because it skips them. In a git repo, git ls-files avoids the exclusion list, because it already leaves out what .gitignore ignores.