float cross2d(vec2 v, vec2 w){ return v.x * w.y - w.x * v.y; } vec2 complex_div(vec2 v, vec2 w){ return vec2(dot(v, w), cross2d(w, v)) / dot(w, w); } vec2 xs_on_clean_parabola(vec2 controls[3]){ /* Given three control points for a quadratic bezier, this returns the two values (x0, x2) such that the section of the parabola y = x^2 between those values is isometric to the given quadratic bezier. Adapated from https://github.com/raphlinus/raphlinus.github.io/blob/master/_posts/2019-12-23-flatten-quadbez.md */ vec2 b0 = controls[0]; vec2 b1 = controls[1]; vec2 b2 = controls[2]; vec2 dd = normalize(2 * b1 - b0 - b2); float u0 = dot(b1 - b0, dd); float u2 = dot(b2 - b1, dd); float cp = cross2d(b2 - b0, dd); return vec2(u0 / cp, u2 / cp); } mat3 map_point_pairs(vec2 src0, vec2 src1, vec2 dest0, vec2 dest1){ /* Returns an orthogonal matrix which will map src0 onto dest0 and src1 onto dest1. */ mat3 shift1 = mat3( 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, -src0.x, -src0.y, 1.0 ); mat3 shift2 = mat3( 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, dest0.x, dest0.y, 1.0 ); // Compute complex division dest_vect / src_vect to determine rotation vec2 complex_rot = complex_div(dest1 - dest0, src1 - src0); mat3 rotate = mat3( complex_rot.x, complex_rot.y, 0.0, -complex_rot.y, complex_rot.x, 0.0, 0.0, 0.0, 1.0 ); return shift2 * rotate * shift1; } mat3 get_xy_to_uv(vec2 controls[3], float bezier_degree){ vec2[2] dest; if (bezier_degree == 1.0){ dest[0] = vec2(0, 0); dest[1] = vec2(1, 0); }else{ vec2 xs = xs_on_clean_parabola(controls); float x0 = xs.x; float x2 = xs.y; dest[0] = vec2(x0, x0 * x0); dest[1] = vec2(x2, x2 * x2); } return map_point_pairs( controls[0], controls[2], dest[0], dest[1] ); } // Orthogonal matrix to convert to a uv space defined so that // b0 goes to [0, 0] and b1 goes to [1, 0] mat4 get_xyz_to_uv(vec3 b0, vec3 b1, vec3 unit_normal){ mat4 shift = mat4( 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, -b0.x, -b0.y, -b0.z, 1 ); float scale_factor = length(b1 - b0); vec3 I = (b1 - b0) / scale_factor; vec3 K = unit_normal; vec3 J = cross(K, I); // Transpose (hence inverse) of matrix taking // i-hat to I, k-hat to unit_normal, and j-hat to their cross mat4 rotate = mat4( I.x, J.x, K.x, 0.0, I.y, J.y, K.y, 0.0, I.z, J.z, K.z, 0.0, 0.0, 0.0, 0.0, 1.0 ); return (1.0 / scale_factor) * rotate * shift; }