Qbasicnews.com

http://www.qbasicnetwork.com/stuff/iregsodu.bmp

Solve this one! I can't figure it out...
Better yet, write an algorithm to solve it!

>anarky

It's unsolvable. I clicked typed in the numbers on sudowin, clicked "solve", and it says "Unsolved".

Do you mean an algorithm that tells you why it is unsolved?...

Hmm, that'd just be some sort of nice picture representation of yetifoot's decision tree. (That is yetifoot's program, right?)


Anyway, first thing I did is put the 4 on position (1, 3) -- (x, y) coords from the top-left position. The 4 is the only thing that goes.

After that, you're stuck. You have to place a 1 in either of the two remaining spots but the positioning of the other "1"s won't allow it.

It's yetifoot's program. It's unsolvable in the traditional sense. I believe we have to work on a distorted "square" where each coloured section nas these properties.

>anarky

One BIG problem. Look at the upper-right square. I won't say more.

I can't believe you haven't seen these before. The traditional 3x3 boxes don't apply - the row and column rules do, but the "boxes" are the coloured regions. I'm assuming the darkened lines are still there just because of the way that program's set up.

Here's a page you might want to check out: http://www.maa.org/editorial/mathgames/m...05_05.html

All you'd have to do is modify the block constraints in yetifoot's program to your custom blocks..

Xerol has it right. The bold lines are left over from the program originally. Ignore them.

I also have one with letters.

>anarky

I've been looking at these sort of sudokus for my next version. I've got super sudoku done, (4x4 A-F), but these type are quite a pain, and will require substantial changes in my engine to make it more generic. Also my grid drawing routines can't handle these types at the moment, and my brain hasn't figured the solution to the grid problem yet. I've still been looking at the 'Dancing Links' algorithm, with no joy yet.

On the bright side though i have done:
4x4 A-F (Super Sudoku)
Speed increases on slower puzzles (thanks in part to DJ Peters for some optimizations, and no thanks to the FB problem with aligning correctly to avoid cache mis-hits)
Better GUI
Hints (For pure logic next steps only)
Ability to choose whether to find all solutions (to check a puzzles full validity)
More generic grid drawing/handling to cope with different types

0.5 coming soon (maybe a month, couple of weeks at earliest)

@anarky, i see you've got sudoku mania as a project in your sig. Are you working on these special puzzles? Any joy? Also if you've looked through my code, and found anything horrible, i'd appreciate any comments.

yetifoot: I am planning Sudoku Maniac as an end user game with randomly generated puzzles. Of course, with your permission, your code will be used to determine the validity of a game, and when a game is deemed valid, the engine will choose a few numbers based on difficulty levels and then get the user to solve it.

If there is more than one solution, all possible solutions will be displayed. As new numbers are filled in, the engine attempt to solve the board to determine how many remaining correct solutions there are.

There are no plans for any other versions yet, since I don't understand how they work.

>anarky

yeah, your free to use that code anyway you want. A credit is appreciated but not necessary.

anyway, heres the solution