Hello everyone!
I’m making a 2D game with a terrain made using marching squares. I’m using the Custom Collider 2D component with edges to “draw” the marching square collider. Here is what a chunk of 25x25 looks like:
The thing is that I want the terrain to be diggable. This means I have to regenerate the collider each time a tile has been dug. The problem is that it’s not performant on the physic side. Calling CustomCollider2D.SetCustomShapes takes 0.2ms and then Physics2D.FindNewContacts takes around 0.5ms. So if I dig between 4 chunks we have a total of 2.8ms which is too much for me.
I tried to reduce the number of shapes in the custom collider but it didn’t change anything. One thing that reduced those spikes is to reduce the chunk’s size. So for example, using 5x5 chunks of 5x5 nodes instead of using 1x1 chunk of 25x25 nodes. But if I want to pool chunks, I’ll want to activate multiple chunks which will also call Physics2D.FindNewContacts when enabled. So the result is the same. Here is me enabling/desactivating a chunk:
I wanted your opinion on my issue:
- Do you think using the custom collider 2d is the way to go? If yes, do you have any idea how to reduce those physic spikes?
- If the colliders in Unity are not made to be edited at runtime, do you think making a custom physic engine for the terrain is worth it?
Thank you for reading 
. As you guessed I’m replacing all shapes inside the CustomCollider2D each time I’m regenerating the chunk. I’ll try to separate the collider into smaller shapes and update only the relevant one.