Hey there, let’s work through that Sudoku Generator problem you’re dealing with! It’s so frustrating when a code snippet you track down online doesn’t behave as expected—trust me, I’ve been in that exact spot.

First, let’s go over common issues that might be breaking the generator implementation you found:

• Unique solvability checks missing: Many generators work by filling a full valid grid then removing numbers, but if the removal step doesn’t verify that each change leaves exactly one valid solution, you’ll end up with grids that have zero or multiple solutions. Double-check if the code uses your working Sudoku Solver to confirm uniqueness after each number removal.
• Backtracking logic flaws: The core generation loop might have bugs in how it selects valid numbers or backtracks when hitting a dead end. Try adding debug logs to track the grid state at each step—look for moments where it gets stuck in an invalid cell or repeats numbers incorrectly.
• Mismatched grid constraints: Some generators have hardcoded assumptions (like only supporting 9x9 grids) that might not align with your project’s setup. Make sure all constants (like box width/height) match what your solver expects.

If fixing that existing code feels like too much of a hassle, here’s a rock-solid alternative approach that leverages your already working solver:

1. Start with a completely empty grid.
2. Fill a random subset of cells with valid numbers, ensuring no duplicates exist in the same row, column, or 3x3 box.
3. Run your Sudoku Solver to fill in the rest of the grid—this guarantees you start with a fully valid, solvable grid.
4. Iteratively remove numbers one by one. After each removal, use your solver to confirm there’s still exactly one unique solution. Stop when you can’t remove any more numbers without breaking that uniqueness.

This method cuts down on validation work since you’re using a tool you already know works. Don’t forget to test edge cases too—try generating grids for different difficulty levels by adjusting how many numbers you leave in, and verify each one with your solver to ensure consistency.

内容的提问来源于stack exchange，提问作者Maxime Michel