A fast, easy to use, free, and community supported 2D game engine
Hi all!!! anyone knows if Cocos2D have a method to get the Color of a pixel?
i need to check that. i see this blog http://www.30daygame.com/2009/04/11/how-the-map-works-or-a-guide-to-platforms/
but doesn't explain how to check the pixel color..
anyone can help me?
Thanks!!!
there was a post earlier on here that links off to the old google groups, with an excellent post by Ernesto.
I'm trying to implement this one myself.
Ernesto's code looks solid, but I am having a hard time understanding a few key points:
Where does the collision image get SET, coordinate wise? I'm using this for a top down game, so the image I'm testing with is a 1024x1024 PNG. The actual GRAPHIC for this world is getting set at the center of my screen to start with, then I am moving my player/camera around over the top of it. The graphic is getting set using a LOCAL coordinate (via setPosition).
But Ernesto's collision map is using a UIImage type, which doesn't like the setPosition code. Where is his map getting set? And is it anchored using it's center point, like sprites and textures?
I assume that I will need to call between my player and the collision map using a WORLD coordinate system to get collision, but I have no idea where the map is at!
The collision image isn't added or positioned anywhere. Its simply read once from left to right, top to bottom I believe. You should probably create an array of the pixel results, so you can check against this during game play.
From memory, I think the image is flipped upside-down compared to the cocos2d co-ordinates. So you'll need to account for that when accessing the relevant Y axis during your comparisons.
So to put it in your terms, it would be at the world coordinate level, with the top left of the map, being the bottom left of your screen.
Hope that helps.
Thanks! I had puzzled that out for myself right before sleep claimed me, but it is great to have confirmation!
Never can be too careful with those late night epiphanies...
Game Smith
Member
Wow, this seems to be a popular topic on the forums lately.
http://www.cocos2d-iphone.org/forum/topic/822
Here is another thread where we seem to be working on the same thing.
Okay, I'm befuddled yet again.
I've basically used Ernesto's magic code from start to finish, with only a few changes to keep the warning and error demons at bay inside of XCode.
Here's the original ERNESTO code, as a refresher:
//1) Load the image:
UIImage *collisionImage = [UIImage imageNamed:@"levelmask.png"];
//2) Get the image size:
int width = (int)collisionImage.size.width;
int height = (int)collisionImage.size.height;
//3) Allocate our array for quick lookups and clear it:
unsigned char *collisionMap = (unsigned char *)malloc( width *
height );
memset( collisionMap, 0, width * height );
//4) Get access to the raw bits of the image
CFDataRef imageData = CGDataProviderCopyData( CGImageGetDataProvider
( collisionImage.CGImage ) );
const UInt32 *pixels = (const UInt32*)CFDataGetBytePtr( imageData );
//5) Build our collision map:
for( j = 0; j < (width * height); j++ )
if ( pixels[j] & 0xff000000 ) collisionMap[j] |= 1;
//6) Release what we no longer need:
CFRelease( imageData );
//Now, whenever you need to see if you hit something, you call a routine
//that does:
-(int)collisionForPoint:(CGPoint)point
{
int x = (int)point.x;
int y = (int)point.y;
// TODO: check for boundaries here
return collisionMap[(y*width)+x];
}I first tried putting the first 6 steps just inside the init of my game's action layer itself. I decided to do this so that we get the data on hand just the once (at init) and don't waste ALL our processing crunches on making and releasing the bloody thing on every update. Then I put the last method further down in the same layer (PAST the init).
It's the last little method that screws me up.
If I place the map maker INSIDE the init, I can't call back to it from the collisionForPoint method, because it tells me that collisionMap (and width) is undeclared.
My fix for this was to make the mapping code its own method, which I then declared an instance of in the @interface of the header, so that I could create the instance in my init and call back to it from my collisionForPoint method.
The problem with this is that I assumed that the maker method would be of the same type that the collisionMap it returns was created: an unsigned char. So my method looked like this:
-(unsigned char) makeCollisionMap
{
UIImage *collisionImage = [UIImage imageNamed:@"bgColl2.png"];
int width = (int)collisionImage.size.width;
int height = (int)collisionImage.size.height;
unsigned char *collisionMap = (unsigned char *)malloc(width * height);
memset(collisionMap, 0, width * height);
CFDataRef imageData = CGDataProviderCopyData(CGImageGetDataProvider(collisionImage.CGImage));
const UInt32 *pixels = (const UInt32*)CFDataGetBytePtr(imageData);
for (int j = 0; j < (width * height); j++)
{
if ( pixels[j] & 0xff000000 ) collisionMap[j] |= 1;
}
CFRelease(imageData);
return collisionMap;
}I initialize this as colly in my init:
colly = [self makeCollisionMap];
Then of course call it from the other method:
-(int) collisionForPoint:(CGPoint)point
{
int a = (int)point.x;
int b = (int)point.y;
int mapValue = colly[(b * 1024)+a];
NSLog(@"val: %i", mapValue);
return mapValue;
}As you can see, I just gave it the constant of my map width (1024) rather than a variable, and then use the NSLog to double check my returned value. And the NSLog gives me a whole crazy range of numbers back from my little array of supposedly 1s and 0s, depending on the player position that I pass into it!
I believe this is because of the way the first method is returning the (unsigned char).
I am getting a warning from Xcode saying:
warning: assignment makes pointer from integer without a cast
at my init.
I assume this is what is ruining my data. I NSLogged my collisionMap as it's being made, and I DID get back my 1s and 0s, so the data INSIDE my collisionMap is good, but it seems to be getting screwed up as I pull it back out of the instance I had to create.
Is anyone still awake after all that? Good! Because I'm ready to kill this silly iMac.
I am open to suggestions/criticism.
Okay, so I guess it's fixed, but I'll be buggered if I know why completely.
After trying all the other Objective-C data types (which sometimes confound me, but THANKS API) I switched my constructor method to data type id. So I have:
-(id) makeCollisionMap
{
UIImage *collisionImage = [UIImage imageNamed:@"bgColl2.png"];
int width = (int)collisionImage.size.width;
int height = (int)collisionImage.size.height;
unsigned char *collisionMap = (unsigned char *)malloc(width * height);
memset(collisionMap, 0, width * height);
CFDataRef imageData = CGDataProviderCopyData(CGImageGetDataProvider(collisionImage.CGImage));
const UInt32 *pixels = (const UInt32*)CFDataGetBytePtr(imageData);
for (int j = 0; j < (width * height); j++)
{
if (pixels[j] & 0xff000000)
{
collisionMap[j] |= 1;
}
}
CFRelease(imageData);
return collisionMap;
}...
I still have my collMap instance declared in the header as an unsigned char, then I had to pass the thing through a DIFFERENT instance name before it all worked... which I don't quite understand, but am happy that it DOES work. Little bobby down there is the "middle-man" variable I make in my init :
unsigned char *bobby = [self makeCollisionMap];
collMap = bobby;I hope this long rant helps somebody else bumbling with Ob-C and cocos2d.
gaminghorror
Member
Instead of:
unsigned char *bobby = [self makeCollisionMap];
collMap = bobby;can't you just write it like this:
collMap = (unsigned char*)[self makeCollisionMap];Maybe you simply forgot the cast. The warning you get seems to indicate that. I bet Objective-C is trying the next best thing and assumes "id" to be a pointer to int.
Anyhow i'm glad i finally found this code on pixel-perfect collisions. It was hard to find so i'll tag my post accordingly.
1 byte per bit? all those wasted bits make me cringe. For those who are learning (etc) try to use the whole byte like so:
to set the map:
int length = width*height;
int byteSize = (length/8)+1; //add padding for length non-divisible by 8
unsigned char *collisionMap = (unsigned char *)malloc(byteSize);
memset(collisionMap, 0, byteSize);
...
for (int j = 0; j < length; j++)
{
if (pixels[j] & 0xff000000)
{
collisionMap[ j / 8 ] |= 1<<(j&7);
}
}
...
to get the flag from the map:
int index = (y*width)+x;
return collisionMap[index>>3] & (1<<(index&7));I haven't tested this, just off the top of my head, but as long as you get the gist. Use 1 bit per pixel for any non-zero pixel.
I guess I come to the right place for my problem in Chipmunk / Cocos2d that I am using for my game.
I am making a game where I want to define a curved floor shape for collision of my player sprite. I was wondering if anyone can answer my question. The function I saw above will return a collision map, but how can I use that (unsigned char) map to define my cpPolyShape in chipmunk? I apologize I am a little beginner in Obj-C and Cocos2d.
Any help would be highly appreciative.
Thanks,
Shafiq
milan
Member
I have 5 screens (512x320) which I scroll in landscape mode. I want to use collision detection method described in this thread.
If my collision picture is 512x320 collision detection works fine for first screen. If I use picture with 1024x320 size (screen 1+2) my collision map is wrong.
Is the method described here supposed to work also on bigger picture than size of a screen?
Only workaround I see is to create 5 pictures for collision detection. Is there better way how to workaround it?
thanks,
Milan
You must log in to post.
