Minesweeper solver

Another take at Minesweeper implemented with Rust. See previous iteration (implemented in Haskell). This implementation does not enumerate consistent layouts - only estimated probabilities are used. This should improve performance by removing combinatorial algorithms compared to robo-minesweeper.

To do

• For immediate tasks see TODO comments.
• API between game and algorithm (supporting batch updates).
• Maybe multi-threaded or GPU version? :)

Bugs

Cannot proceed when surrounded

Sample

3 % # # # # # # 
% % # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 


Scratch
3 @ # # # # # # 
@ @ # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # # 
# # # # # # # #  

Edge {}
thread 'main' panicked at 'No position selected.', src/main.rs:379:25
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Should select something at random in unexplored area

Done

• Removed border padding - it did not solve the problem of edge conditions completely.
• Convert storage to arrays, sparse implementation is not strictly needed as we can use mutation now. For optimal presentation would need both sparse and regular presentations. Sparse representation is more complex (e.g. hash or 2d tree), so putting it aside until really needed.
• Consider rust-ndarray/ndarray to process field.