rebar3_erli18n_jaro (rebar3_erli18n v0.1.0)

Copy Markdown View Source

Jaro-similarity fuzzy matcher for msgmerge-style merge.

When merge finds a msgid in the old catalog that no longer appears in the freshly extracted .pot, it tries to pair it with a NEW msgid (one present in the new .pot but absent from the old catalog) so the translator's work can carry over as a #, fuzzy entry instead of being lost to #~ obsolete. The pairing uses string:jaro_similarity/2 (stdlib, OTP 27+), comparing each removed msgid against each added msgid.

A pair is accepted only when its similarity is at or above the threshold (default 0.8, matching GNU msgmerge's fuzzy heuristic spirit). Among candidates above the threshold the highest score wins; ties break deterministically on the candidate's position in the supplied list (earlier wins), so the result never depends on map iteration order.

The comparison is bounded: O(|removed| x |added|) similarity calls, each over two bounded strings — there is no catalog cross-product beyond that.

Summary

Functions

best_match(Needle, Candidates)

Find the best fuzzy match for Needle among Candidates, default threshold.

best_match(Needle, Candidates, Threshold)

Find the best fuzzy match for Needle among Candidates at Threshold.

similarity(A, B)

Jaro similarity of two binaries, in 0.0..1.0.

Functions

best_match(Needle, Candidates)

-spec best_match(binary(), [binary()]) -> {ok, binary(), float()} | nomatch.

Find the best fuzzy match for Needle among Candidates, default threshold.

Equivalent to best_match(Needle, Candidates, 0.8).

best_match(Needle, Candidates, Threshold)

-spec best_match(binary(), [binary()], float()) -> {ok, binary(), float()} | nomatch.

Find the best fuzzy match for Needle among Candidates at Threshold.

Returns {ok, Match, Score} for the highest-scoring candidate whose similarity is >= Threshold, breaking ties toward the earlier candidate in the list. Returns nomatch when no candidate reaches the threshold (or the list is empty).

similarity(A, B)

-spec similarity(binary(), binary()) -> float().

Jaro similarity of two binaries, in 0.0..1.0.

A thin wrapper over string:jaro_similarity/2 that accepts binaries directly. Two empty strings are defined as fully similar (1.0), matching the stdlib.