Lodestar.Extensions.MathNet 0.1.0. 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.

MathNetInterop.ToSparseMatrix

Converts a CsrMatrix into a Math.NET SparseMatrix.

public static SparseMatrix ToSparseMatrix(CsrMatrix matrix)

Parameters — matrix is the matrix to convert.

Returns — a SparseMatrix of the same shape holding the same values, sharing no array with the source.

Exceptions — ArgumentNullException when matrix is null.

Example — a row whose columns arrive out of order still reads correctly.

using Lodestar.Abstractions;
using Lodestar.Extensions.MathNet;
using MathNet.Numerics.LinearAlgebra.Double;

var unsorted = new CsrMatrix(
    rowCount: 1,
    columnCount: 5,
    values: [9.0, 7.0, 8.0],
    columnIndices: [4, 0, 2],
    rowPointers: [0, 3]);

SparseMatrix sparse = MathNetInterop.ToSparseMatrix(unsorted);

double first = sparse[0, 0];    // => 7
double middle = sparse[0, 2];   // => 8
double last = sparse[0, 4];     // => 9
double gap = sparse[0, 3];      // => 0

Remarks — the rows are sorted, and duplicate columns are added together. That is not tidiness: CsrMatrix never promised an order, Math.NET reaches a cell by searching the row, and a matrix handed over unsorted would convert without complaint and then answer lookups with zeros. The example above is exactly that case.

The cost is a single pass that decides whether anything needs doing. Everything this repository's vectorizers produce is already sorted, so the common path is one comparison per stored value and no allocation beyond the copy Math.NET makes anyway. Only a matrix built by hand pays for the sort.

Explicit zeros are carried across rather than dropped — Math.NET's own compressed-row storage counts "stored values including explicit zeros", so the structure survives unchanged.

Applies to — net10.0, netstandard2.0.

See also — MathNetInterop, MathNetInterop.ToCsrMatrix.