This is an excercise for personal instruction to build a sudoku solver in Haskell. No guaranees can be made for efficiency or effectiveness.
The idea is to build a Monad Stack which sets up a series of rules and constraints - each of which can maintain it's own state. Any level of the stack can cause the game to fail (by evaluating to mzero/empty). They communicate by filtering the total number of available options.
Currently winning is managed by a seperate function - but it would be nice to manage this in the type system in future.
I also intend to use this to learn a bit more about testing frameworks and documentation. Currently it is a bit of a mess.
I have been testing this on GHC8.2. An earlier version was failing to build due to a bug in GHC8.02, so mileage on lesser versions may vary.
If you have the nix package manager, a nix-build or nix-shell command should
be adequate - the derivations read from the cabal file.
Currently, there are five test files. Run the solver on them by:
git clone https://github.com/atp30/sudok.git
cd sudok
cabal build
sudoku < tests/test1.txt
sudoku < tests/test2.txt
sudoku < tests/very_hard.txt
Currently, test1.txt and test2.txt are solved entirely deterministically.
very_hard.txt requires trial and error. I am uncertain about the others,
although it seems as though some trial and error may be used, based on times.
test1.txt and test2.txt are sourced from sudopedia and licensed under the GNU FDL.
the wiki* tests are sourced from wikipedia