Blackbox: is a clever two-player deductive game invented by Eric Solomon in the 1970s and is traditionally my first testing-ground for any new programming language. In Java it proved a good introduction to event handling, array handling, layout managers, random number generation and a few other things besides. Perhaps, above all, it taught me never to touch GridBagLayout again.

Tile puzzles: This is an original home-grown puzzle, first published in the Lubenham Village Newsletter (circa 1982). One of it's most striking features is that it isn't a maze (not sure how that happened). The original 10-piece tile set (introduced December '97) was new at that time, and was refined and tested to ensure only a single solution.

Tile puzzle required reading several new chapters on Java, including (several times over) one particularly baffling section on how to control screen flicker. Image handling was an easier section, but unfortunately also turned out to be the wrong section, graphics handling would have been better since the initial version of tile puzzle used forty gif files and could be very slow to load. Tile puzzle was thus belatedly revamped in February '98 to draw it's own graphics.

Knight's tour: The knight's tour is an old idea with a new twist. I first experimented with mazes based on the knight's move as a child but this was the first to have a special moving 'peep-hole' through which the maze is viewed. The peep-hole principle could be applied to any maze of course, but the knight's tour also offered a test-bed for some simple Java graphics.

Implementing the knight's tour maze went remarkably smoothly (for a change). It actually took longer to come up with a satisfactory maze than write the Java. That said, I found fillRect() rather troublesome; is it meant to include or exclude the width of the border? I found I had to both draw and fill each stepping stone to achieve consistent results across all browsers.

Tilt mazes: [☕] The tilt mazes were the first new concept I explored with clickmazes specifically in mind. At the time (1997) they were my first new mazes for some years. Perversely, the more complex 3D tilt mazes came first, it was only after implementation that I decided the tilt behaviour was sufficiently interesting in isolation to merit the development of a simple 2D version.

The tilt mazes as originally launched, required little or no new knowledge of Java. I had a little trouble initially getting the main panel (container) to redraw itself with a different set of components (yes, yes... validate() - so obvious!?). It did take somewhat longer than a Sunday afternoon this time, but then so did the mazes.

In April '98 the tilt mazes were significantly reworked to make use of the new base classes introduced by the tilt puzzles (see below).

In October '99 the 2D tilt mazes were the subject of further major rework. Many more mazes added (various sizes) and a range of brand new multi-goal mazes were added. The Java diverged here for this development (unfortunately) but perhaps one day I'll bring it all back together again.

Tilt puzzles: [☕] The tilt puzzles take the tilt behaviour one step further and turn the basic 2D tilt maze into something more of a novelty. Two sliding blocks now share the same tray and must reach their own separate goals without coming into contact. Superficially the puzzles are just two mini mazes in one, the challenge however is the synchronisation.

The tilt puzzle Java could have been a simple variation on what had come before, but the desire to implement a realistic slide effect forced me to dabble with something new... Java threads. Then, inevitably, my one small runnable class not so much slid but snowballed into a major overhaul of all the tilt-maze source. The tilt puzzles subsequently provided the base classes from which all tilt-maze classes were derived. For an OO-impaired software engineer this turned out to be quite a voyage of discovery.

Colour-zone mazes: This is a clickmazes original variation on the rolling block puzzle concept.

The Java here was nothing new (I shrugged off yet again the pressure to leap to Java 2) and thus took only a few hours to implement. The mazes however took a fair bit longer to devise and, even so, turned out limited in complexity, there's just a little bit too much connectivity with no other rules to restrict movement.

Marble mazes: The marble tilt mazes were my next foray into the tilt-maze concept. This is the code I should have written for the original tilt-puzzles (March '98) but found a little beyond me at the time. Hence the original 'bang' effect as the blocks slid into each other (which was more or less what my code did, the first time I ran it). Some years later, and with a good deal more practice at thread and vector handling behind me, things came together with slightly less explosive results.

Oskar's hysteresis mazes: The hysteresis mazes are inspired by an original mechanical puzzle by Oskar van Deventer. Oskar kindly invited me to try implementing his puzzle in Java, and this worked very well indeed, although somehow the resulting mazes turned out not quite devious enough for either of us. So we increased the size, threw in a black-hole or two and finally sprinkled everything with strange yellow dots. Consider the yellow dots 'portals' into a strange inner dimension; the curious north–south maze, that lurks behind every east–west maze.

Maze of life: [☕] The maze of life is a puzzle based on John Conway's well known Game of Life algorithm. An interactive Game of Life no less! So hey, we should all be experts. However this 'game' is about as unpredictable and uncontrollable as life itself. Basic survival is soon mastered, but are you a true master of your fate? Can you reach that ultimate goal? Fortunately in this version you always have the undo option (oh how useful that would be!).

Maze of life V2 was added in July 2000 and offers a means of saving and recording played sequences. Further enhancements were added in April 2001 (V2.2) to provide unlimited redo and preloading of size, start state and move sequence. In Jan 2002 (V2.4) the 'Undo all' option was added plus several fixes to support much larger move sequences and play areas.

Plank puzzles: This puzzle has become one of the best known and most accoladed members of the clickmazes puzzle family. Famously designed on the back-of-an-envelope, while waiting in the A&E dept of the Luton and Dunstable NHS hospital (it was a long wait and I only had the contents of a hand-bag to amuse myself).

The plank puzzle applet is the work of Graham Rogers and is an honorary member of the clickmazes applet family. I also thank one Andy Williams who was the first to develop a successful plank-puzzle solver.

Oskar's dragon maze: [☕] The dragon maze is an ingenious fractal maze design by Oskar van Deventer. Use the applet to drag the layers of interwoven lattice-work one at a time from left to right and watch the maze develop. The secret is the white areas of each layer are transparent, allowing underlying colours to show through.

This applet required a small diversion back to the theory books to discover how to create an image with transparent areas. In fact you don't (well I didn't) you set up a filter instead. Nothing is ever quite what it seems.

No-left-turn maze: Looks trivial until your realise you can't turn left (or make U-turns). Some old Java here, but I thought it about time it went on-line (maze No.2 was added new).

Oskar's Four-bit mazes: Java2 at last (or should that be Java 1.3.0?). That wasn't so bad after all and it seems even GridBag layout can be mastered, with sufficient time and patience.

I needed a small, non-graphical, challenge to test my new found skills. Oskar's four-bit mazes were perfect. If you manage to switch off all those lights just be careful not to fall over the furniture.

HexaRoll: [☕] Another great puzzle challenge from Oskar van Deventer, based on a variation of the tilt theme. But oh, applets with hexagons and rotating shapes?... that has to be one for Graham. With lots of late night development consultancy from Louie the result was sure to be another perfectly executed applet.

So what exactly was my contribution? Well I did find the odd bug, and even tried to fix a couple, but my main task was to encode up the mazes, which as you might see from the HTML source took a little while to sort out. And then having gone to all that trouble I really couldn't then resist adding a couple more mazes of my own.

Two more exotic variations on the tilt theme made possible by a multi-purpose applet developed by Graham which supported just about all forms of tilt-puzzle dabbled with so far and a few more yet to be invented. So this list will probably grow and grow. So far we have...

The magnetic-block puzzles: [☕] These puzzles were inspired by another web-app popular at the time called Denki. Denki did not itself use the tilt-rule but adding it allowed creation of some really tiny, meany puzzles. Try them and see, or try something less mind-blowing...

The cup and pea puzzles: which were invented when Graham mis-understood an idea suggested by Andy Williams. A fortuitous accident that yielded the most exotic tilt-variation so far.

Orientation mazes: [☕] Ooops I'm definitely slipping behind here... yet another applet by a fellow designer and this time the puzzles aren't my own either. Well at least I thought up the concept. Nice applet by Jimmy Stephens of Atlanta, including elegant graphics. Puzzles are so tiny and yet so very, very mean, all thanks to OrientExpress (an extension of SwampBeast).

BoxUp: At last some new java all of my own. A sort of evolution of Graham's magblock code, but with all the standard stuff separated off into its own class (puzzlePresenter) to enable rapid development of new puzzles. The clever bit is puzzlePresenter dynamically loads a second puzzle-specific class (boxup in this case) which must implement the stdpuzz interface. Took a little while to manage all the funny looking exceptions that can occur if things go wrong.

An alternative name for this is Inside-out Sokoban so if you like Sokoban try this.

Wriggle puzzles: [☕] OK, I'm on a roll now. Actually it can't be a roll - I've done that already, so it must a wriggle. My puzzlePresenter classes are making this all very easy, but I keep adding new interfaces to allow optional elements of game play. This one required the introduction of a interface to handle mouse motion.

Looks like another collection is underway here... Colour-wriggle mazes added August '06, TJ-wriggle puzzles added August '07 and Extreme TJ-wriggle added December '10.

Oskar's collection: [☕] Oskar meet java... java meet Oskar. Ignoring an early misfortunate detour into the deep dead-end of Swing (all-singing, all dancing - but very browser unfriendly) there's no doubt Oskar has taken firm command of his very own virtual puzzle world (and a four-dimensional, parallel-universe world at that).

Polyfold puzzles: [☕] Another Oskar creation reworked and extended out of almost all recognition for the web. Initially implemented using some basic trig functions, but which later proved superfluous. The biggest hurdle was implementing my own version of the Line2D.intersects() method. Not so much referencing the JAVA text-books here, it was more back to my O-level maths notes.

Chain-reaction: [☕] This java was simple enough except for the auto-generation feature where it took a while to fine tune the algorithm such that the search for a new puzzle had an acceptably low number of solutions, yet didn't take for-ever to find them. Then there was the need to revisit java Threads, the first time for a while. Thanks to the auto-generation feature this is the first of my creations with some replay value.

Full-house puzzles: This is where puzzlePresenter, and all that came before and after, really starts to pay-off. A couple of hours tinkering with some core logic is all it takes to knock out something new. My thanks to Erich Friedman for dreaming up such a clever and elegant puzzle, and for lending me a few samples to present here.

XL-up maze: Oh to be only 40 again, it seems like so long ago now. The hardest challenge with this one was trying to write my own mini 'beasty' program to eek out the sequence as far as possible. I'm still hoping to outlive the best score.

Step-over sequence mazes: Another quick cookie-cut from the puzzlePresenter cake of giving. Prototyping the full-size physical walk-round version took eons longer. But 15 years later the second prototype is still going strong!

Quantum maze: No faffing about with complicated puzzle encodings here, each puzzle can be encoded in just a few bits. With this puzzle the start-state is a blank canvas, quite literally.

Stepping-stone mazes: A surprisingly accommodating maze with almost perfect connectivity and very low risk of black-holes. So much so that I called this the self-wiring multi-state maze when I presented a talk on it at G4G12. The 5-minute presentation took a lot longer to prepare than the java, and was a lot more stressful.

Puff-ball puzzles: Shook the dust of the beasty code again for this one, to hunt for those deepest-state goals. If the solution eludes you (as they do mostly for me!) take a peek at the HTML source for the web-page, you'll find the solution hidden there, in plain sight.