Adding NonRecursiveNegamax - Ready to Merge

[JohnAD]

This is not ready for merger. It is simply a point of progress for this project.

There is, in fact, some kind of fundamental flaw in my NR version.

The test suites run a game of Knights for each different AI (see file ai_unittests.py). The regular and NR versions should choose the same moves. And they do until deep into the game: move 18. So there is some border case that is being handled differently. Can't for the life of me see why. I suspect my version is being more aggressive in the alpha/beta pruning; but that is a wild guess at this point.

Next step: run an end-game by hand on paper. Get exact answer. Compare against both. It might take me a while...

[JohnAD]

@Zulko

Hooray! Finished debugging the code! Still don't merge though. I need to update documentation first. I will add more comments as well.

BTW, do you want the version to be '0.0.6' or do you want take this to a stable release number, such as '1.0.0'?

BTW, I did modify the original 'Negamax' slightly. See

https://github.com/Zulko/easyAI/pull/21/files#diff-74bb0325c1491d187c46b5fd3b7e1b0f

When debugging on paper I was able to determine that sometimes, in fairly rare border cases, the function's returned bestValue would not correspond with the best_move that has that value. It seems to mostly affect leaf nodes so it very rarely produces a wrong answer.

There is also now a ai_unittest.py script in the root directory. Running it will do tests against all of the AI's to verify integrity.

[JohnAD]

Finished! Testing works and I've updated the documentation.

Two items to point out:

1. I updated version to 1.0.0.1 from 0.0.0.5. If you don't want me to do that, I can make it 0.0.0.6. IMO, this is a stable library.

2. I'm not 100% sure of my description of DUAL in the documentation changes. See https://github.com/PurpleSquirrelGames/easyAI/blob/master/docs/source/ai_descriptions.rst

Let me know if you would like to see additional changes (or want me to undo some of the changes I made.) Otherwise, it is all ready to merge.

[JohnAD]

@Zulko

I'll have quite a bit of time this weekend to work on this. So, if you have a chance, let me know if there are additional changes you would like me to make/unmake. :)

John

[JohnAD]

Since the PR had not yet been accepted, I went ahead added another commit that handles the unusual case of asking for a depth of 0 of the NonRecursiveNegamax. Strictly speaking it really isn't negamax anymore at depth zero. So, it is not a critical fix at all; but it is good to be thorough, IMO.