Solving Sudoku like a human
A heuristic solver that never backtracks
The standard computer-science answer to Sudoku is backtracking: pick a branch, and if it contradicts, unwind. It works, and it is nothing like how a person solves a puzzle. I like to solve mine in one pass.
So this solver does what a human does. Every cell tracks its own candidate set, and a digit goes down only when it is the sole option for that cell, or that cell is the sole option for the digit. No branching, no unwinding — the constraints propagate until the grid falls out.
It’s built around a single Cell class holding a position, a value, and a candidate set. Once that exists and the candidate sets stay honest, the puzzle mostly solves itself. Written on a long flight, as an excuse to apply decent software engineering to a puzzle I’ve liked since I was a kid.
