Lodestar.Fuzzy 0.5.0. This page is frozen at that release. Read the current documentation for what main says now. A link to a decision or a migration page follows main, and leaves the archive.

Home › Fuzzy matching

A score for every pair of two collections — what Process.Cdist answers with.

public readonly struct ScoreMatrix

Properties — Rows is how many queries were scored, Columns how many choices each was scored against.

Example — the shape, one cell, and one row.

using Lodestar.Fuzzy;

string[] queries = ["new york", "boston"];
string[] choices = ["new york mets", "boston red sox", "atlanta braves"];

ScoreMatrix scores = Process.Cdist(queries, choices);

int cells = scores.Rows * scores.Columns;              // => 6
double first = Math.Round(scores[0, 0], 4);            // => 76.1905

ReadOnlySpan<double> boston = scores.Row(1);
double bostonBest = Math.Round(boston[1], 4);          // => 60

Remarks — a type rather than a bare double[], because the shape is not an input here. Every bulk member of Lodestar.Preprocessing takes a featureCount beside its span, so a caller already holds the width; Cdist derives it from the two lists, and handing back an array alone would make the caller carry the column count next to it.

The storage is row-major, the layout every bulk member in this repository uses: cell [row, column] lives at row * Columns + column, and ToArray hands it over in that order. Row is a window onto it and copies nothing.

Equality is on the storage, not the cells. Two matrices are equal when they are the same matrix; comparing two sets of scores element-wise is what Row is for. That keeps Equals O(1) on a type whose whole point is to be large.

Applies to — net10.0, netstandard2.0.

See also — Process.Cdist, the matching index.