Questionable vector equality?

The following is the Equals method of Vertex3f:

/// <summary>
/// Indicates whether the this IVertex3 is equal to another IVertex3.
/// </summary>
/// <param name="other">
/// An IVertex3 to compare with this object.
/// </param>
/// <returns>
/// It returns true if the this IVertex3 is equal to <paramref name="other"/>; otherwise, false.
/// </returns>
public bool Equals(IVertex3 other)
{
	const float Epsilon = 1e-6f;

	if (ReferenceEquals(null, other))
		return false;

	if (Math.Abs(X - other.X) >= Epsilon)
		return (false);
	if (Math.Abs(Y - other.Y) >= Epsilon)
		return (false);
	if (Math.Abs(Z - other.Z) >= Epsilon)
		return (false);

	return (true);
}

I do not think that using an absolute Epsilon is the right approach in Equals(). This can lead to strange behaviors. Consider the following code:

Vertex3f v1 = new Vertex3f(1e-7f);
Vertex3f v2 = new Vertex3f(1e-9f);
bool equal = (v1 == v2); // = true

I suggest that you get rid of epsilon entirely and stick to true equality (especially since the summary of the method indicates that it is a true equality).

You could add an “approximately equal” method but in that case use a relative error. Something like: Math.Abs(X - other.X) >= Epsilon*Math.Max(Math.Abs(X), Math.Abs(other.X))

Your OpenGL binding is very good. Many thanks for your impressive work!

[antoinecla]

On a related subject, what do you think about adding static Translation, Rotation, etc? This would allow constructs such as: shaderProgram.SetUniform(ctx, name, Matrix4x4.Rotation(q)*Matrix4x4.Translation(translation));

Instead of:

transform = new ModelMatrix();
transform.Translate(translation);
transform.Rotate(q);
shaderProgram.SetUniform(ctx, name, transform);

[luca-piccioni]

Vertex* tests requires some more testing, but the skeleton is done. Matrix classes are on the way, but surely they definitively need a review and a test set.