Text osa distance

Development build. This page describes main, not a released package. The latest published Lodestar.Text is 0.4.0 — read its documentation.

Osa.Distance

Counts the fewest insertions, deletions, substitutions and swaps of neighbouring characters, with no character allowed to take part in more than one edit.

public static int Distance(ReadOnlySpan<char> a, ReadOnlySpan<char> b, TextElement element = TextElement.Utf16Unit)
public static int Distance<T>(ReadOnlySpan<T> a, ReadOnlySpan<T> b) where T : IEquatable<T>

Parameters — a and b are the two strings to compare. element says what counts as one character: TextElement.Utf16Unit by default, or TextElement.CodePoint for rapidfuzz's answer outside the Basic Multilingual Plane. The second overload compares any two spans of an IEquatable<T>.

Returns — int, the number of edits. Zero when the two are equal, and never negative.

Example — the pair that separates OSA from full Damerau-Levenshtein, which answers 2.

using Lodestar.Text.Distances;

int d = Osa.Distance("CA", "ABC");   // => 3

Remarks — for real text this and DamerauLevenshtein agree almost always, and this one costs less to compute — three rolling rows instead of a full matrix and a symbol table. Reach for it as the default transposition-aware distance, and only move to DamerauLevenshtein if the pairs you are matching really do need a stretch edited twice.

The trap is that "almost always" is not always, and the disagreement is silent. "CA" to "ABC" is 2 under DamerauLevenshtein and 3 here, because reaching 2 means transposing CA to AC and then inserting into that same stretch. If a test suite was built against Python's DamerauLevenshtein.distance, Osa.Distance will pass on nearly every case and fail on a handful, which is the worst way to discover the difference.

The restriction costs one property outright: unlike Levenshtein and unlike unrestricted DamerauLevenshtein, this is not a metric. The triangle inequality fails — Osa.Distance("bca", "ab") is 3, while going through "ba" costs 1 + 1 — so a BK-tree or any other structure that assumes a metric will silently return wrong neighbours. Use DamerauLevenshtein when you need to index rather than to score.

Applies to — net10.0, netstandard2.0.

See also — Osa.NormalizedSimilarity, DamerauLevenshtein.Distance, Levenshtein.Distance, the Python equivalence table.