The New Sudoku Players' Forum

[horybori]

Hi all,

First of all, I'm glad to publish my sudoku solver here for the first time. My program differs from every other solver that I have been already known. It uses an universal algorithm to solve "every" popular variations of sudoku.
It can solve sudokus with...
- irregular units/regions: x-sudoku, hyper-sudoku/windoku, jigsaw, nonomino...
- multiple boards: like samurai sudokus
- cages: e.g. killer sudokus (every four operations are supported: addition, multiplication, division, and subtraction)
- relations between cages: greater, less, equal relations (supported, but user cannot set yet)
- additional rules: greater than sudokus, consecutive sudokus
- additional restrictions: custom candidate set can be specified for every cell (e.g. even or odd numbers only)
- custom symbols: wordoku
...in every size (from mini to giant and more...).


I'm looking forward to your feedback.


P.S.: English isn’t my first language, so please excuse any mistakes and feel free to send translation corrections too. I plan to make a usable documentation about every little feature if it will be needed.

[surendra.jain]

Does your solver uses brute force algorithm for solving all these kind of puzzles or it solves the puzzles by logic. I have written codes to solve sudoku, killer sudoku, iaognal sudoku, irregular sudoku, samurai sudoku, kakuro, calcudoku etc.. I have also developed a software that solves sudoku puzzles from pure logic. The source code can be downloaded from my webite :

sites.google.com/site/skjgeek

Surendra

[horybori]

It's basically a simple backtrack algorithm. It uses naked/hidden single techniques only because advanced techniques can't be used efficiently. They would be very costly (both implement and use) compared to the number of examined candidates. Instead it tries to guess which is faster and more effective in the vast majority.

[tarek]

My program differs from every other solver that I have been already known. It uses an universal algorithm to solve "every" popular variations of sudoku.

Welcome to the forum,

How fast does your solver prove that this Kazaguruma WIndoku X has 1 solution so:
1. Kazaguruma (Windmill)
2. Each board is Windoku X (Hyper-Sudoku X) (4 windows & Diagonals)

This was posted in another thread with an image but the image no longer exists

Code: Select all

..................4...8........3................5.......6......5.................
.2....3.............8.........1..........91......................................
.................................................................................
......................................75..........3.........1.............4....7.
.................9......7.......5................8........3...7..................

Tarek

[horybori]

I haven't heard about this puzzle variation yet, but I managed to found an example here: http://forum.enjoysudoku.com/tarek-s-windoku-kazaguruma-windmill-x-t30283.html

(Intel® Core™ i5-2410M Processor (3M Cache, up to 2.90 GHz), 8 GB RAM, Windows 10)

Results:
- first solution only: 76 ms
- with uniqueness check: 472 ms

I also solved the other sudoku and I got similar results (94 ms and 610 ms).

I attached the puzzle so you can try it out for yourself easily.

[tarek]

Great. The puzzle I posted for you to test is actually puzzle 3 in that thread. It is the same puzzle variant as puzzle 1

The 5 line representation is to make it easy to load, share and store these puzzles. My suggestion would be to implement this easy feature in your solver as it more or less universal between solvers.

Puzzle 3 is a Machine challenging puzzle. So it was a good benchmark to test your solver!!!

Another puzzle to test is the Unsolvable killer sudoku #41 from the list I posted in this thread http://www.rcbroughton.co.uk/sudoku/forum/viewtopic.php?f=3&t=434

Tarek

[horybori]

I have a good news and a bad news. The good is that I finally managed to solve the puzzle after ~1 hour computing, but the bad news is that the puzzle isn't unique.

My first two (wrong) solutions (repeating values in cages):

Code: Select all

953216874718594326264837159491672538876359412325481967189725643637148295542963781
953126874718594326264783159491672538876359412325841967189265743637418295542937681


[tarek]

I have a good news and a bad news. The good is that I finally managed to solve the puzzle after ~1 hour computing, but the bad news is that the puzzle isn't unique.

I agree and disagree .... I probably should have mentioned it but in most cases it is implied:

The original published Killer sudokus in the Times (& that is why most players follow that convention) had also a sum-cage constraint where you can't repeat a symbol in the cage & this was the reason most players know the number combinations based on that. So unless specified; the convention is that there should be no repeated symbols in the sum cages.

This again can be modified easily in your solver to accept both possibilities but you should set the default to: not accepting symbol repetitions in cages.

Tarek

[horybori]

I'm here again. I modified my solver as you suggested.
Furthermore, I've implemented some advanced cage solving techniques, so it's a little bit faster now.

Solution of the unsolvable #41 (without repeating symbols):

Code: Select all

283197546967542813415368729591726384876439152324851967149275638752683491638914275

Hidden Text: Show

puzzle.png (18.9 KiB) Viewed 3424 times

[tarek]

Excellent. Well done!!! This was designed to test machine solving so you can use it also to test the timings following any improvements in your solver in the future.

Your programming path is as follows:

1. Solving
2. Generating
3. Rating/analyzing

Good luck,

Tarek