Description:
According to the official Rapier JS docs
intersectionsWithShape should provide more precise results compared to collidersWithAabbIntersectingAabb.
I tested both functions and confirmed that intersectionsWithShape indeed gives more accurate intersection information.
However, when reviewing the source code (world.ts), both functions appear to use the broad-phase system. This seems inconsistent with the observed accuracy difference.
Expected Behavior:
intersectionsWithShape should operate in the narrow phase (as implied by its accuracy).
Actual Behavior:
Both functions appear to rely on the broad phase in the source, yet intersectionsWithShape behaves as if it performs narrow-phase checks.
Question:
Why does intersectionsWithShape behave more accurately if both functions are implemented in the broad phase?