Depth first search
**About this Kata**
Good for teaching: recursion
Related Katas: KataBreadthFirstSearch
Depth-first search is not just an AI technique it is also the standard way that trees (ubiquitous in computer science) are traversed.
For this (easiest) form of depth-first search, steer towards using the function-call stack, rather than a stack data structure.
Try to avoid implementing a Graph or Maze or Problem class (even as a mocked interface) for your depth-first-searcher to traverse. Instead, ask questions of a “human” user (e.g. standard in/standard out or whatever your language uses). For example, “What are the exits?” and “Where are we?”. This allows exploratory testing, which is fun. Having to build a Graph or Maze by initializing data, or parsing a file, slows down the group and is distant from the main thrust of the kata.
Of course, you should mock the conversation in your unit tests! This demonstrates that exploratory testing can be recorded in repeatable automated tests, which is valuable.
Suggested Test Cases
The one-node graph. The two-node graph. A 2x2 maze. A full two-level binary tree. A 3x3 maze.
Comments from those who are working on this Kata
If you have more time and a moderately advanced group, after getting this working, add a requirement of “event driven”. That is, the interface to your solver should be something like “Step”, which asks at most one question of the Graph/Maze /Problem /stdin object, and then returns. This models requirements change. The changes to your test cases should be relatively mild, but the changes to the internals of the algorithm might be fairly drastic.