Evaluation of an expression dealing with Vector

[anouarlachheb]

Hi,
I am using MathNet.Symbolics .Net Nuget package v 0.24.0

I am trying to evaluate an expression containing Vector variable.
i.e:
<Tlin_Vector = (Tref_Vector - 25) * (2)^(16)> - (Tref_Vector - 25) * (2)^(16)>

I am constructing dynamically (by C#) the variables dictionary.
i.e:

        // create a dense zero-vector of length 8
        var PRef = Vector<double>.Build.Dense(9); // 9 points are needed

        // create a dense zero-vector of length 8
        var TLin = Vector<double>.Build.Dense(9); // 9 points are needed

and then assign them to the dictionary.

My problem is that I catch an exception:
goto IL_0166;
IL_0066:
switch (u.Tag)
{
default:
switch (v.Tag)
{
default:
throw new Exception("not supported");

[diluculo]

I'm experiencing the same issue. Here's a minimal example:

// This fails with "not supported"
var expr = SymbolicExpression.Parse("2*s"); // 2 * ComplexVector
var sVector = Vector<Complex>.Build.DenseOfEnumerable(
    frequencies.Select(f => new Complex(0, 2 * Math.PI * f))
);
var variables = new Dictionary<string, FloatingPoint> {
    ["s"] = FloatingPoint.NewComplexVector(sVector)
};
var result = expr.Evaluate(variables);

Looking at the current implementation in Evaluate.fs (both v0.24.0 and v0.25.0):

let fmultiply u v =
    match u, v with
    | Real x, Real y -> Real (x*y)
    | Complex x, Complex y -> Complex (x*y)
    | RealVector x, RealVector y -> Real (x*y)        // Vector dot product only
    | ComplexVector x, ComplexVector y -> Complex (x*y) // Vector dot product only
    | RealMatrix x, RealMatrix y -> RealMatrix (x*y)  // Matrix multiplication only
    // No scalar-vector mixed operations

It seems like scalar-vector element-wise operations are missing. This would be extremely valuable for vectorized symbolic evaluation.

If this wasn't intentional, it would be great to have support for these operations (scalar-vector, scalar-matrix and so on) added. Similar patterns would probably need to be added to other functions like fadd, fpower, etc. as well.