Feedforward control plays a key role in achieving high performance for industrial motion systems that perform non-repeating motion tasks. Recently, learning techniques have been proposed to further improve both performance and robustness to non-repeating tasks by using a rational feedforward basis. The aim of this paper is to propose a unifying framework which connects these approaches. Experimental results on an industrial motion system validate the approaches and illustrate benefits of rational feedforward tuning in motion systems, including pre- and post-actuation through stable inversion.