We consider electric vehicle routing problems which occur in emerging transportation and logistics contexts: last-mile delivery with light vehicles, industrial regional operations, and service robotics. We formulate the problem as a set-partitioning model with path-based variables, and solve the linear relaxation with column generation. Our main contribution is a subpath-based decomposition of the pricing problem which happens in two phases: we first generate subpaths between charging stations, and then stitch them together to form full paths. We formulate a formal domination framework based on subpaths and paths, and show that it results in a correct pricing problem algorithm. Within this framework, we then design the right resources, resource extension functions (REFs), and domination criteria that solve the pricing problem for various EVRP problem elements (load, time windows, and nonlinear charging) and vehicle-routing relaxation tightening strategies (ng-routes, subset-row cuts). Computational experiments demonstrate that we outperform traditional path-based labelling benchmarks, especially when there are enough subpaths per path and enough customers/tasks per subpath.(An earlier version of this work was recognized as one of four finalists of the INFORMS TSL Best Student Paper Award in 2024!)