lrbridge / mango Goto Github PK

I wrote test cases to expose the remaining TODOs in A* and Greedy. I'm in the process of fixing but haven't quite gotten it done yet.

• if we come across the same state again, replace node in frontier with lower cost node
• make tiebreakers better for A* (use greedy, than closer x, than closer y, than up/down/left/right convention?)

We should change heuristics from all being calculated up front to just the heuristic from the state you are in being calculated at run time.

Then, we should update the TurnsHeuristic to be smarter since it now knows what direction it is facing for sure at a given square.

Part 1 is implemented, but maybe we should each take one of the four larger sample mazes and see if the solution looks right for each of the 4 algorithms?

I've gone through many of the smaller mazes and they seem to be right, but I've never gone through the solutions on these mazes:

• mediumMaze.txt
• bigMaze.txt
• openMaze.txt

Sound worthwhile? Does anyone want to sign up for testing one of those mazes?

We need to pick one of the optional pieces. Thoughts on which?