I’m looking for beta participants for my PC client! See http://blog.sc.tri-bit.com/archives/151.

My new puzzle test and markup engine (5^TH generation!) now supports quite a few types, and I’m going to be supporting more types later.

Probably the coolest thing about my current setup is that I can mix and match rulesets, layouts and renderer types at will, or if I want to, even at random. This means the paint metaphor actually supports a whole bunch of different puzzle types now. (Hacking them into the interface has been double-painul.)

The current types I support:

• Normal sudoku at any N×M size (includes 4×4 Super Sudoku and 5×5 Sudoku the Giant, plus much larger)
• Sudoku-X at any N×M size
• Killer Sudoku at any N² size
• Killer Sudoku-X at any N² size
• Nonomino Sudoku (AKA Irregular Sudoku, Geometry Number Place or Kikagaku Nanpure) at any N² size
• LCDoku Sudoku / Digital Magic Number sudoku puzzles at any N×M size
• Disjoint Set Sudoku / Non-Connected Nonomino Sudoku at any N² size
• Hyperdoku Sudoku / 4D Sudoku
• Hyperdoku-X Sudoku / 4D Sudoku X

• TwoDoku Sudoku / Sensei / Gattai-2 sudoku puzzles at any N² size
• DoubleDoku Sudoku / Gattai Near-2 sudoku puzzles at any N² size
• Triple X Sudoku / Gattai Half-Near-3 sudoku puzzles at any N² odd-n size
• Ring Sudoku / Gattai Near-4 sudoku puzzles at any N² size
• Samurai Sudoku / Gattai-5 sudoku puzzles at any N² size
• Flower Sudoku / Gattai Near-5 sudoku puzzles at any N² even-N size
• Wing X Sudoku / Gattai Half-Near-5 sudoku puzzles at any N² odd-N size
• Butterfly Sudoku / Gattai Far-5 sudoku puzzles at any N² even-N size
• Shogun Sudoku / Harakiri Sudoku / Gattai-11 sudoku puzzles at any N² size
• Daimyo Sudoku / Sumo Sudoku / Gattai-13 sudoku puzzles at any N² size
• Shaolin Sudoku / Gattai-25 sudoku puzzles at any N² size

• Arbitrary Non-2d Surfaces
• Non-square Grid / Alternate Tesselation shapes

• Numeric grid renderer
• AlphaDoku grid renderer
• Alphanumeric grid renderer
• Wordoku Sudoku / Scramblets grid renderer
• Color patch grid renderer

Things I intend to support soon, or things whose current implementations are unsatisfactory:

• Binary Background Sudoku (If the predicate is evenness, these are called Even/Odd sudoku or Guusuu Kisuu Nanpure; if the predicate is region count and there are no more than one given for each digit, these are called Magic Sudoku)
• Magic Sudoku (Warning: PDF)
• GT/LT Sudoku
  
• DominoDoku / Domino Sudoku
  
• Disallowed Number Sudoku / Hitotsu Chigai Nanpure
• Color Killer Sudoku – my own variant, similar in several ways to Disallowed Number but with non-summing killer range boxes; very, very difficult
• Color Killer GT-X Sudoku (really)
• Clueless Sudoku

• Template pre-form markup

• Image patch grid renderer
• CodeDoku Sudoku grid renderer
• Binary Background grid renderer
• Disjoint Set and Hyperdoku grid renderer
• Sequential Sudoku Renderer

Things I’m not yet sure whether I’ll support:

• Metadoku
• ChessDoku
• Latin Squares

Anyway, if you see something I’m missing, please let me know. I am fibbing a little bit in this list: killer works fine in my old renderer, but I haven’t brought it over to the new one yet. I’m still using the old renderer to make those; there are serious problems with the way I handle killer in the old layout engine, as the interface just paints it as a layer on top of a layer. However, since the new renderer doesn’t actually know anything about layout (the user inputs it) it can’t handle Killer adjacency the way it should. I’ll fix it, but I haven’t yet, and I’m not entirely sure how just yet. Getting GT/LT puzzles working was a bear and a half.

On the bright side, I’m back down to about two minutes to paint up a new Sudoku puzzle with five rulesets on at the same time; I can paint them faster than people using pure computer generators do just to make them. Yay for understanding machine cache behavior. DLX is a great algorithm, if you don’t give a damn about cache performance, but for the same reason judy trees are superior to normal trees despite being inoptimal due to hardware concerns, my negascout ply sudoku test driver massively outperforms any of the programs I’ve seen so far with five rulesets on, compared to their one.

In other words: pwnt.