what you have here is a graph.  you've figured out how to draw it, and now you want to model it.  that's the opposite thinking of what i'd suggest:  first figure out how to model it, then figure out how to render the model - much simpler in the end.

 

having said that, here's one way.  btw, this isn't the way that I'd do it, if building from scratch I'd have a graph with nodes and edges to render "symbolically" as tiles based on their degree/valence.  but the following method may "plug in" easier to what you already have, as you can just store int's in an array along with your tile data...

 

you could probably represent the internal contents of each cell with 5 bits like so:

int CONNECT_RIGHT = 0x1;

int CONNECT_UP = 0x2;

int CONNECT_LEFT = 0x4;

int CONNECT_DOWN = 0x8;

int SINKSOURCE = 0x10;

 

for each edge that's "connected", that bit is 1, giving you all 16 possibilities.  if a cell is completely empty, all bits are zero, value is zero.  if a cell is one of your "big dots" it'd also have the SINKSOURCE bit on (these are your "clickable" cells).

 

determining "connectedness" to neighboring cells goes like this:

 

if my bit in a certain direction is on,

  get the neighboring cell in that direction

  if neighbor's bit for the opposing direction is on

    then the two of us are connected

    (otherwise we have a "half connection" which is probably invalid)

 

note the possibility of modelling invalid states - that wouldn't happen with a true graph model, but just requires a bit of care on your part when building the model.

 

start by clearing an "on" flag for all cells (i'd use the same set of bitmasked values), then set that flag for the clicked cell.  for a graph this small, a simple recursive function might work,  it would check the current cell, mark it if "on", then recurse for each of its four neighbors that's connected (and not already marked on, so you don't backtrack infinitely!).   redraw, and just color each "edge" based on its "on" flag, and viola!, you've just reinvented depth-first pathfinding!  :D

I use this library in our library navigator to find the shortest path between two points. I will put the source code on our web server and create a new topic in the Exhibition forum.

Hi

 

davbol: gosh, all this time what I thought was a graph is (according to wiki) really a "graph of a function" or a "chart". It'll take me some time to understand this new world of vertices and edges. I understandish everything else you've suggested though.

 

amnon.owed: the tiles are created procedurally, and I've separated out common elements.

 

mbraendle: I've had a look at that library, but think that I need to have a go at reinventing the wheel first so that I understand what's going on behind the scenes.

 

Don't hold your breath for its exhibition!

thanks!

string