Is it possible to make it so that the Rectangles game will have a slightly wider area you can click on for it to think you have clicked on an edge? I keep not being able to undo them again. (I know zip about making computers do this sort of thing as yet, so if this sounds like an absolutely daft request I do apologise.)

No, that's a perfectly reasonable request, and in fact I already moved that value back and forth several times because I wasn't entirely sure where it should optimally be :-)

<f/x: checks> Hmm. It currently seems to be set at 0.3, which means that 30% of the width of a square on each side of the line is treated as edge, and the remaining 40% is centre. That's actually quite a lot already, so I'm unsure about making it any bigger.

Is it possible (he says, clutching at straws) that what's biting you is not the narrowness of the region that counts as an edge, but actually the largeness of the region that counts as a corner (and therefore the program isn't sure which edge you meant)? If that were the case then I could improve matters by lowering that 0.3 figure. Ideally I should arrange for it to be configurable and let people tweak it back and forth until they're happier...

Now I think about it, I do find myself pressing U a lot (handy keyboard shortcut to Undo). Perhaps a rethink of the control mechanism is in order.

Feedback.

Highlight the edge that would be clicked on if the user clicked, as the mouse hovers. If it wouldn't count as an edge by the click filter, don't.

Sadly, one of the platforms I'm hoping to (eventually) have this collection ported to is stylus-based PDAs, so any control mechanism that depends on mouse hover events for its effectiveness is a bad plan...

Well, the control/feedback mechanism doesn't *have* to be identical on all platforms surely?

You could also subtly shade active areas so they're visually distinct (without interfering with the larger distinctions on the board.)

I'd like the control mechanism to work similarly on all platforms, because that way adding a new platform front end is O(1) effort rather than O(number of games), and adding a new game back end is O(1) rather than O(number of platforms). Since my ultimate aim is to end up with loads of games all running on loads of platforms, this will eventually become a serious concern :-)

Anyway, I think I've improved matters now without having to resort to mouse hovering.

Surely the Right Thing is for the sensitive area for each edge to be a square at 45˚ centred on the middle of the edge?

Unfortunately I also want to be able to drag out an entire rectangle, and also importantly I want to be able to drag out a line (for when you know almost exactly where a rectangle goes and want to mark in as much of its edge as is common to all its possible positions, and don't want to do that by clicking ten edge pieces individually). For these purposes, you don't always want to start your drag at an edge; starting at centres or corners is useful as well.

Therefore, I've been subdividing each square into a 3x3 grid, with four "corner" sections, one "centre" section, and four "edge" sections. I've been indecisively fiddling with the parameter that states how close to the edges the dividing lines are, but the problem is that as you increase that parameter then the edge region gets shorter as it gets fatter, and vice versa.

I've now decided it might be better to be able to independently vary the sizes of the sensitive square around each corner and centre. So now the diagram looks more like a square with a small square at each corner, a larger square in the middle, and diagonal lines in between; so the edge region is octagonal, and its width and height can be varied independently.

Preliminary alpha testing suggests that this is much better all round; once I've tinkered a bit more I'll check it in and see what other people think.

That might have been it. Try it and see.

Actually, after a bit more playing with it I decided to go for something rather more sophisticated. Rather than having a simple 3x3 grid dividing each square into corner, edge and centre regions (left), I went for a more complex model (right) which expands the edge regions at the expense of both corners and centres:



That seems to me to be generally better all round. The version of Rectangles now on my website has been updated to do this; give it a try and let me know if it's any better?

Oh, that's definitely better. Much easier to use.

And I love the way it now does a little flash when you complete the puzzle!

*grin* Two people actually asked for the flash - they'd got used to it in some of the other puzzles and missed it in Rectangles. I thought this one was already obvious enough when you'd completed it that I needn't bother, but no, it seems once you give people one flash they're addicted :-)