Normalization ready
Downcase and strip @ sigils from input and every candidate, so @ivar and IVar compare alike — the first step of MRI’s SpellChecker#correct.
go-ruby-did-you-mean is a pure-Go (no cgo) reimplementation of the matcher at the heart of Ruby's did_you_mean standard library — MRI's DidYouMean::SpellChecker#correct and the two string-distance metrics it ranks with, Jaro–Winkler and Levenshtein. Given a mistyped word and a dictionary it returns Ruby's exact ranked list of spelling suggestions — a faithful, byte-for-byte port of upstream spell_checker.rb, jaro_winkler.rb and levenshtein.rb, without any Ruby runtime. It is a standalone, reusable module — a sibling of go-ruby-yaml, go-ruby-regexp and go-ruby-erb — and the did_you_mean matcher for go-embedded-ruby.
Jaro–Winkler filter & ranking ready
Keep candidates whose Jaro–Winkler similarity clears the threshold (0.834 for inputs longer than 3 characters, else 0.77), drop an exact match of the original, and rank survivors by similarity descending — a stable sort then reverse, matching Ruby’s tie order.
Levenshtein mistype filter ready
Keep the ranked candidates within an edit distance of ⌈len/4⌉ of the input, exactly as MRI’s mistype filter does.
Misspell fallback ready
If nothing survives, return the first ranked candidate whose Levenshtein distance is strictly less than the shorter of the two word lengths — Ruby’s stable-sorted, input-order fallback.
Unicode-correct metrics ready
Jaro / JaroWinkler.distance and Levenshtein.distance ported exactly, comparing Unicode code points (not bytes) like Ruby’s String#each_codepoint, so multibyte words (café, naïve) rank identically.
Differential oracle & coverage ready
Golden vectors captured from MRI plus a broad corpus of typos, transpositions, case variants, multibyte words and no-match inputs fed to the system ruby’s SpellChecker#correct, reproduced byte-for-byte. 100% coverage, gofmt + go vet clean, green across all six 64-bit Go arches and three OSes.
- ready capability complete (exit criteria met, nothing material left)
- active in progress (substantial work shipped, not yet finished)
- planned not started yet (or downstream by design)
A faithful, line-for-line port of MRI's SpellChecker#correct in pure Go, cgo disabled, so it cross-compiles and embeds anywhere. It normalizes input and candidates, applies the Jaro–Winkler similarity filter and ranking, the Levenshtein mistype filter, and the misspell fallback — comparing Unicode code points (not bytes) like Ruby's String#each_codepoint, so multibyte words rank identically. Validated differentially against the system ruby's DidYouMean::SpellChecker#correct. The did_you_mean matcher for the sibling org github.com/go-embedded-ruby.