On the inversion of computable functions