.

https://github.com/MoonPresident/Latin-Square-Generator

Calling "generate_latin_square" will generate a solution, and calling "cull_latin_square" will generate a problem. Passing the solution into "get_tower_clues" will give you the hints that surround the problem.

Incidentally, I have updated the code to go up to 15x15.

If your solution for the 6x6 would also pass the 7x7, you have a solution that is too fast for this kata.

"should" and "would" are matter of style, and therefore personal preference.

Don't know if I'm doing it wrong, but the way I wrote it, I should theoretically directly apply 7 to my function and it would solve that as well. I mean do any principles actually change by the size?

I can confirm it can be solved.

Excellent. I will try it out soon. Im working right now on an more efficient solution. I will try it with the old program which could solve in time until it reached 10x10 and the newer potentially faster one.

Hi Sandro, I have changed the problem set. Its more diverse but also potentially more difficult, however it should have removed the particular problems that were causing you difficulty. If you have another go, please let me know how you find it.

I will also double check it again. Maybe it was an error in the program.

Hi Sandro, I can confirm that these test cases have a single valid solution using my algorithm.

Sounds great. I could find 2 11by11 Puzzles somewhere in the internet and used them for my unit tests.

Yes I read about that. In the meantime I discussed my solution with some people and we came up that it is not that efficient yet. I try now an completly different approach to solve the puzzles faster. We will see If thats enough for the tests.

Hi Sandro. The 11 by 11's are fairly computationaly expensive. I am intending to optimise the backend of this puzzle to give the frontend more time to run, if you read down through this post you will see that there are some problems uploading C++ vectors that I need to work out.