codingben / box2d-netstandard

C# port of Box2D - Erin Catto's 2D Physics Engine
https://box2d.org
MIT License
263 stars 53 forks source link

Breaking solid objects in half #42

Open MadCowChicken opened 2 years ago

MadCowChicken commented 2 years ago

Description

This isn't actually an issue, I'm just looking for guidance.

Suppose I have a rectangle (representing a brick) and a circle (representing a slow-ish bullet) and if the bullet hits the brick with enough speed I want to turn the brick from one object into two or three objects along fracture lines that seem intuitive.

Do you think it would work for me to modify the ContactManager.Collide() method in such a way that it does two passes... The first pass could replace the bottom part of the function that looks like

// Here we destroy contacts that cease to overlap in the broad-phase.
if (overlap == false)
{
    Contact cNuke = c;
    c = cNuke.GetNext();
    Destroy(cNuke);
    continue;
}

// The contact persists.
c.Update(m_contactListener);
c = c.GetNext();

with something like

if (overlap == true)
{
    ...break brick into pieces if conditions are met...
}

and then the second pass would be exactly the code that's in the ContactManager.Collide() method normally?

Or would this not work? Maybe I would need to rewind positions after the first pass? Possibly by capturing the world state at the beginning of the first pass and resetting it to that (except for the broken brick) before doing the second pass?

thomasvt commented 2 years ago

Hello,

Splitting a brick into pieces is game logic, and not the job of the physics system (box2d). The job of the physics system is to simulate interaction of physical bodies that the game logic defined earlier. The Physics engine should not be used to decide when to change or create new bodies based on game-specific rules like the one you are describing.

Therefore I suggest you do not change any internal code of box2d, but to use box2d's ContactListener (many tutorials online. eg. https://www.iforce2d.net/b2dtut/collision-callbacks (should be trivial to translate to C# equivalent).

Your ContactListener implementation will receive collision notifications from box2d. You can then check if one body is a bullet and the other body is a brick. Splitting the brick will work best by deleting the brick entirely and creating several new brickfragments as separate game objects (each with their own RigidBody).

By the way: This kind of question is more a working-with-box2d question, or even general gamedev question. You will reach a lot more people by asking things like this through eg. StackOverflow. The original box2d is C++, so many people will respond from that C++ perspective, but the box2d concepts translate perfectly to C# with Box2d-netstandard. So, you should be able to get help from them too.

codingben commented 2 years ago

I agree with @thomasvt. @MadCowChicken Feel free to see examples how to use ContactFilter in related issue.

MadCowChicken commented 2 years ago

I didn't know it had ContactListener / ContactFilter sort of support. Thank you.

MadCowChicken commented 2 years ago

Actually a quick look at the collision callbacks link above makes me believe it is not what I'm looking for. I don't need to add code to run after a collision is processed. I need to add code to run before a collision is processed.

I need to split a brick the moment before the bullet hits the brick so that the pieces can fly apart more naturally.

thomasvt commented 2 years ago

So, you want the ball to hit the brick, break it and have the same impact velocity propagate realistically to the broken pieces?

I don't think you can create that without faking some things here and there. Some ideas:

MadCowChicken commented 2 years ago

If I create a copy of the bullet with IsSensor = true and fix it to the original, is there a way to guarantee that the sensor collisions will happen before the bullet collisions? And also allow my dynamic fracturing code to be run before the bullet collisions, rather than running my dynamic fracturing code at the end of the collision process?