quote:Original post by DudeMiester
Just my 2 cents:
I imagine it that the objects ask the world to check if they are colliding with anything, and they pass it a sort of outline of itself. The world then passes this iformation to any nearby objects and asks them if they collide with this outline. If anything does, then the world sends the object back where, when and with what to the object, so it can handle the situation accordingly.
Big problem with this method. If your character is moving 10 tiles per second. There's a wall in front of him that is 1 tile thick. .11 seconds pass between updates. Your character would be .1 tiles in front of the wall, so he would be able to walk through walls. Given enough of a system "lag" he could walk through any wall.
My solution was to define a minimum distance. .05 tiles for example. Now you actually calculate every frame of minimum distance between where the character was last frame and where his velocity would take him this turn. So if he was at x = 1 last frame, and ends up at x = 2 next frame.. there wouldn't be 1 collision test, but 20 called.. one at 1.05, 1.1, 1.15, etc, etc..
This effectively solves the problem as long as the minimum distance is greater than the minimum sized obstacle. Working in whole numbers is alot easier too.. but I wanted to have tiles that could be 1/2 walkable and 1/2 non-walkable.
If you try to implement a perfectly OO collision detection system its going to probably be awkward at best. Don't let one overhyped paradigm rule over your coding. OO is a tool, not a rule. I would really love to know of any commerical game that was written with strict adherance to 'classical' OO. I really don't think its possible. As a side note I don't really consider singletons and other concepts that aren't OO but people like to pretend they are, as OO. Shrink wrapped globals are still globals.
[edited by - haro on May 28, 2003 12:23:00 PM]
).
another approach would be to implement collision detection somewhere outside, so the map-data can be fetched from the room, the positions from the NPCs etc., do all the collsion detection and "inform" the NPCs accordingly. but this approach again would have the drawback that a lot of functionality is taken away from the objects, so it''s not really encapsulated anymore...
now feel free to give me some hints on how to do this 
