Using greedy algorithms to generate de Bruijn sequences is a classical approach that has produced numerous interesting theoretical results. This paper investigates an algorithm which we call the Generalized Prefer-Opposite (GPO). It includes all prior greedy algorithms, with the exception of the Fleury Algorithm applied on the de Bruijn graph, as specific instances. The GPO Algorithm can produce any binary periodic sequences with nonlinear complexity at least two on input a pair of suitable feedback function and initial state. In particular, a sufficient and necessary condition for the GPO Algorithm to generate binary de Bruijn sequences is established. This requires the use of feedback functions with a unique cycle or loop in their respective state graphs. Moreover, we discuss modifications to the GPO Algorithm to handle more families of feedback functions whose state graphs have multiple cycles or loops. These culminate in a graph joining method. Several large classes of feedback functions are subsequently used to illustrate how the GPO Algorithm and its modification into the Graph Joining Prefer-Opposite (GJPO) Algorithm work in practice.