vectormath library and matrix operations

Question

I found this copy on github to link to but I am using the one downloaded from sourceforge.

My question about the way they design their matrix operations seems really strange to me.

For example, if I create a 4×4 matrix and set it's scale. Then I would like to rotate that previous matrix using the vectormath matrix library seems to reset the matrix back to an identity matrix then applies the rotation which doesn't make any sense why this would happen.

Take a look at this function to make a rotation

static inline void vmathM4MakeRotationY( VmathMatrix4 *result, float radians )
{
    float s, c;
    s = sinf( radians );
    c = cosf( radians );
    vmathV4MakeFromElems( &result->col0, c, 0.0f, -s, 0.0f );
    vmathV4MakeYAxis( &result->col1 );
    vmathV4MakeFromElems( &result->col2, s, 0.0f, c, 0.0f );
    vmathV4MakeWAxis( &result->col3 );
}

Does this library expect you to keep multiple matrices around use one to apply rotations then multiply?

edit

this is some previous matrix math code that I used to rotate a matrix, it looks like this.

mat4_s mat4_rotateX(mat4_s* out, float angle, mat4_s* inMat)
{
    float s = sinf(angle),
            c = cosf(angle),
            a10 = inMat->m[4],
            a11 = inMat->m[5],
            a12 = inMat->m[6],
            a13 = inMat->m[7],
            a20 = inMat->m[8],
            a21 = inMat->m[9],
            a22 = inMat->m[10],
            a23 = inMat->m[11];

    if (!out->m) {
        for(size_t i = 0; i < 16; i++)
        {
            out->m[i] = inMat->m[i];
        }
    } else if (inMat->m != out->m) { // If the source and destination differ, copy the unchanged rows
        out->m[0] = inMat->m[0];
        out->m[1] = inMat->m[1];
        out->m[2] = inMat->m[2];
        out->m[3] = inMat->m[3];

        out->m[12] = inMat->m[12];
        out->m[13] = inMat->m[13];
        out->m[14] = inMat->m[14];
        out->m[15] = inMat->m[15];
    }


    out->m[4] = a10 * c + a20 * s;
    out->m[5] = a11 * c + a21 * s;
    out->m[6] = a12 * c + a22 * s;
    out->m[7] = a13 * c + a23 * s;

    out->m[8] = a10 * -s + a20 * c;
    out->m[9] = a11 * -s + a21 * c;
    out->m[10] = a12 * -s + a22 * c;
    out->m[11] = a13 * -s + a23 * c;
    return *out;
}

this is the process that I have to take to get vectormath to do the same thing.

mat4* mat4_rotate_y(mat4* out, const float angle){
    mat4 m;
    mat4_identity_v(&m);
    vmathM4MakeRotationY(m, angle);
    mat4_multi(out, out, m);
    return out;
}

the multiplication code is fairly standard but the vmathM4MakeRotationY looks like this:

static inline void vmathM4MakeRotationZ( VmathMatrix4 *result, float radians )
{
    float s, c;
    s = sinf( radians );
    c = cosf( radians );
    vmathV4MakeFromElems( &result->col0, c, s, 0.0f, 0.0f );
    vmathV4MakeFromElems( &result->col1, -s, c, 0.0f, 0.0f );
    vmathV4MakeZAxis( &result->col2 );
    vmathV4MakeWAxis( &result->col3 );
}

just for completeness the vmathV4Make_Axis looks like this:

static inline void vmathV4MakeZAxis(VmathVector4 *result) {
    vmathV4MakeFromElems(result, 0.0f, 0.0f, 1.0f, 0.0f);
}

vmathV4MakeFromElms looks like this:

static inline void vmathV4MakeFromElems(VmathVector4 *result, float _x,
                                        float _y, float _z, float _w) {
    result->x = _x;
    result->y = _y;
    result->z = _z;
    result->w = _w;
}

Answer 1

This function seems to do "Initialize the matrix to be a rotation transform" instead of what you are expecting, "Add a rotation transform to the current transform".

Like you say, you can work around by storing the rotation in a separate temporary matrix and then multiplying:

VmathMatrix4 temp;
vmathM4MakeRotationY(&temp, 1.23);
vmathM4Mul(&mytransform, &mytransform, &temp);

That is pretty much what a hypothetical vmathM4ApplyRotationY() would have to do anyway.