[fansenze]

Summary

• defer autofix and suggestion construction until the corresponding edit demand requests it
• use a validated fixed-width range for the ASCII any keyword, with the scanner-backed range helper as a safe fallback
• reuse the validated range for plain unknown and never replacements while preserving the parent replacement for keyof any
• add demand-sensitive, range, fix, suggestion, trivia, and disable-directive coverage

Go benchmark

Package-local benchmarks were run with benchmem under stable idle conditions. This comparison isolates the final validated keyword-range and range-reuse optimization by comparing the deferred-report intermediate implementation with the final implementation.

Case	Time before -> after	Delta	Bytes before -> after	Allocs before -> after
Diagnostics only	6665.5 -> 1399.7 ns/op	-79.0%	10632 -> 392 B/op	68 -> 4
No matches	234.3 -> 230.4 ns/op	-1.7%	392 -> 392 B/op	4 -> 4
Suggestions	25171.6 -> 9432.6 ns/op	-62.5%	43912 -> 13192 B/op	452 -> 260
Fix diagnostics only	6705.5 -> 1392.4 ns/op	-79.2%	10632 -> 392 B/op	68 -> 4
Autofix	14480.9 -> 4089.2 ns/op	-71.8%	23944 -> 3464 B/op	260 -> 132
Ignored rest args	495.4 -> 495.4 ns/op	0.0%	392 -> 392 B/op	4 -> 4
keyof any suggestions	16914.4 -> 12409.6 ns/op	-26.6%	28040 -> 17800 B/op	324 -> 260

Validation completed:

• go test ./internal/plugins/typescript/rules/no_explicit_any -count=3
• pnpm run check-spell
• pnpm run format:check
• changed-package golangci-lint with --new-from-rev=origin/main
• git diff --check

Related Links

N/A

Checklist

• Tests updated (or not required).
• Documentation updated (or not required).