mefisto94 / jme3-recast4j Goto Github PK

Whenever FormationHandler#target is null, because there was no call to requestMoveToTarget(), trying to move any actor individually leads to an exception (at least when the FormationHandler like the CircleFormationHandler aren't prepared for that).

There would've been multiple solutions for this (e.g. reworking FormationHandlers, makeing the caller more clever), but I am thinking about removing Formations completely or if not redesigning the way they are handled completely.

My reasoning for this is: While moving multiple agents to the same location WILL cause trouble with physics, when stopping right in time, the agents don't collide (when coming from multiple directions). Apart from that, this requires domain specific knowledge (how and where should the agents be placed, sometimes it depends on where they come from, many applications don't even use physics) while at the same time introducing complexity into the library and removing control from the caller.

The code currently is quite complicated with a few if's, some proximity detections and others. This state is unobservable by the caller and there is no good way to influence it.

See MeFisto94/jme3-recast4j-demo#11 and https://github.com/MeFisto94/jme3-recast4j-demo/blob/65f39b7104dbb4b171745b46619017cfd1a5c001/src/main/java/com/jme3/recast4j/demo/states/NavState.java#L248