2015-01-19-1453-EST

Mon Jan 19 14:53:37 EST 2015

# journal - Attempting to make some conceptual progress
Going to fill out a bug report idea using the theory from the last entry.

Going to rubber duck right here some of the ways I see this bug can be solved.
Can use this stream of conscience data to build a more thorough bug fix idea.

One possible idea is that I should be rewrite the player collision to be LESS stateful.
To look at the problem with more of a functional hammer.
With this input, step forward in time and produce a new version of everything.

What happens if there is a circle of dependent collisions?

E    A --> B
     ^     |
     |     v
     D <-- C

In all other cases, the dependence of collisions could be dealt with using recursive descent.
But in the case above how to I avoid infinite looping.
Is this collision even solvable?

What happens if I am provided both the current state and the potential future?
Is this enough information to break the infinite loop? No, I don't think it is.

### pseudo code for an algorithm to break this loop

```
A is moving into B.
Does B have any unique collisions it must Solve? Yes
Are we here because of this unsolved collision? No

B is moving into C.
Does C have any unique collisions is must Solve? Yes
Are we here because of this unsolved collision? No

C is moving into D.
Does D have any unique collisions it must Solve? Yes
Are we here because of this unsolved collision? No

D is moving into A.
Does A have any unique collisions it must Solve? Yes
Are we here because of this unsolved collision? Yes
Then we must solve `D is moving into A` with the information we have.
    - We could make blanket assumption that D cannot move.
        This means that D will lose a movement race from E.
        At first glance, I think this would be fine.
```

If the above algorithm will work then I should use recursive descent.
To do this I need to start storing the collisions per actor from the broad phase.
This is necessary so the questions about dependency can be asked.

What happens if the recursion starts at B instead of A?
It is non deterministic because then A would be the one denied a potentially valid move.
I think this would be fine for now.
Would this cause any problems?
Leave it up to tests to decide if this system is alright.

Mon Jan 19 16:23:41 EST 2015