We present a novel methodological framework for quantum spatial search, generalising the Childs & Goldstone (CG) algorithm via alternating applications of marked-vertex phase shifts and continuous-time quantum walks. We determine closed form expressions for the optimal walk time and phase shift parameters for periodic graphs. These parameters are chosen to rotate the system about subsets of the graph Laplacian eigenstates, amplifying the probability of measuring the marked vertex. The state evolution is asymptotically optimal for any class of periodic graphs having a fixed number of unique eigenvalues. We demonstrate the effectiveness of the algorithm by applying it to obtain O(vN) search on a variety of graphs. One important class is the n × n3 rook graph, which has N = n4 vertices. On this class of graphs the CG algorithm performs suboptimally, achieving only O(N-1/8) overlap after time O(N5/8). Using the new alternating phase-walk framework, we show that O(1) overlap is obtained in O(vN) phase-walk iterations.