In my last question seen here: Sudoku – Region testing I asked how to check the 3×3 regions and someone was able to give me a satisfactory answer (although it involved a LOT of tinkering to get it working how I wanted to, since they didn’t mention what the class table_t was.)
I finished the project and was able to create a sudoku generator, but it feels like it’s contrived. And I feel like I’ve somehow overcomplicated things by taking a very brute-force approach to generating the puzzles.
Essentially my goal is to create a 9×9 grid with 9- 3×3 regions. Each row / col / region must use the numbers 1-9 only once.
The way that I went about solving this was by using a 2-dimensional array to place numbers at random, 3 rows at a time. Once the 3 rows were done it would check the 3 rows, and 3 regions and each vertical col up to the 3rd position. As it iterated through it would do the same until the array was filled, but due to the fact that I was filling with rand, and checking each row / column / region multiple times it felt very inefficient.
Is there an “easier” way to go about doing this with any type of data construct aside from a 2d array? Is there an easier way to check each 3×3 region that might coincide with checking either vert or horizontal better? From a standpoint of computation I can’t see too many ways to do it more efficiently without swelling the size of the code dramatically.
I built a sudoku game a while ago and used the dancing links algorithm by Donald Knuth to generate the puzzles. I found these sites very helpful in learning and implementing the algorithm
http://en.wikipedia.org/wiki/Dancing_Links
http://cgi.cse.unsw.edu.au/~xche635/dlx_sodoku/
http://garethrees.org/2007/06/10/zendoku-generation/