The concept of recursion was easiest to grasp by watching a kid do the maze on a placemat at a restaurant. You go in a line until you find a dead end, then backtrack. Repeat the process until you reach the goal. Each intersection where you make a decision is your function call, and each backtrack is the return result.
Which is an odd example, since maze solving by recursion typically requires something like the Zipper data structure. And... no, that is not necessarily easier to deal with than just using a standard stack and jotting down your work as you go.
The pseudocode will search the maze, find the exit, then print its location followed by the path to get there in reverse order. Nothing complicated, but shows off the statefulness of recursion without needing anything else.
And this is one where I find myself falling back onto (ironically, per the thread?) Dykstra's algorithm. Which I did not learn recursively. It makes much more sense because the "path" that I am recording is a first class thing in the algorithm and not a byproduct of the implementation.
