Box2D coordinate system concepts: PTM_RATIO and retina display

Question

Recently I published a question regarding this topic and I received a useful answer, but my experimentation points me in a different way that I don’t understamd.

From the answer is clear that we should use the same PTM_RATIO for retina and non-retina devices. However we may double it from iPhone to iPad if we want to show the same portion of the world. In my case I used 50 for iPhone and 100 for iPad because Box2d simulations works better if the bodies are between 0.1 and 10m and the main sprite is about 2m.

I used Physics Editor to build the fixtures using GB2ShapeCache without success for retina devices. Then I decided to feed Box2D coordinates directly and I reached strange conclusions that I would like to clarify.

I created a debug method (independent from any sprite) to draw a single line from 1/3 of screens height to 1/3 of screens wide.

- (void)debugGround
{
    // iPad: 1024x768
    // iPhone: 480x320
    CGSize winSize = [CCDirector sharedDirector].winSize;    // unit is points
    b2EdgeShape groundShape;
    b2FixtureDef groundFixtureDef;
    groundFixtureDef.shape = &groundShape;
    groundFixtureDef.density = 0.0;

    b2Vec2 left = b2Vec2(0, winSize.height/3/PTM_RATIO);
    b2Vec2 right = b2Vec2(winSize.width/3/PTM_RATIO, winSize.height/3/PTM_RATIO);
    groundShape.Set(left, right);
    groundBody->CreateFixture(&groundFixtureDef);
}

If Box2D takes coordinates in points and converts them dividing by PTM_RATIO, the result should be the same for iPhone and iPad retina and non retina.

The result for iPad non retina is as expected:

But for iPhone retina and iPad retina, the fixtures are doubled!!

The most obvious correction should be divide by 2, this means dividing by CC_CONTENT_SCALE_FACTOR.

I managed to make it work for all devices refactoring the code to:

- (void)debugGround
{
    CGSize winSize = [CCDirector sharedDirector].winSize;
    b2EdgeShape groundShape;
    b2FixtureDef groundFixtureDef;
    groundFixtureDef.shape = &groundShape;
    groundFixtureDef.density = 0.0;

    b2Vec2 left = b2Vec2(0, winSize.height/3/PTM_RATIO/CC_CONTENT_SCALE_FACTOR());
    b2Vec2 right = b2Vec2(winSize.width/3/PTM_RATIO/CC_CONTENT_SCALE_FACTOR(), winSize.height/3/PTM_RATIO/CC_CONTENT_SCALE_FACTOR());

    groundShape.Set(left, right);
    groundBody->CreateFixture(&groundFixtureDef);
}

I also managed to display correctly the lower platforms dividing by the scale the vertex, the offsets and anywhere I use PTM_RATIO to convert to Box2D coordinates.

It is supposed I shouldn’t use CC_CONTENT_SCALE_FACTOR by any means to multiply positions because GL functions already take this into consideration.

Can anyone clarify this behavior? In which concepts I’m wrong?

I hope this helps the community to understand better Box2D coordinate system.

Answer 1

you misunderstood: GL functions (this includes ccDraw* functions!) require multiplication with content scale factor because GL works on pixel resolution, whereas UIKit views and cocos2d nodes use point coordinates.