Batch createMultiBody returns non-unique ids after removeBody

[rmsc]

I'm periodically creating batches of few hundreds of objects with createMultiBody, and also periodically remove a few of them using removeBody. After removing objects from the world, some of the ids returned by crateMultiBody are actually colliding with ids of existing objects.

I've noticed that the returned ids are contiguous and seem to start at len(existing_bodies). This obviously breaks if random objects are removed.

This is pybullet 3.05.

[rmsc]

Indeed that seems to be the case:

int uid = b3GetStatusBodyIndex(statusHandle);
if (numBatchPositions > 0)
{
        PyObject* pyResultList = PyTuple_New(numBatchPositions );
        for (i = 0; i < numBatchPositions; i++)
        {
                PyTuple_SetItem(pyResultList, i, PyLong_FromLong(uid - numBatchPositions + i + 1));
        }
        return pyResultList;
}
else
{
        PyObject* ob = PyLong_FromLong(uid);
        return ob;
}

[erwincoumans]

Thanks for the report. Can you create a small repro case/test?

[rmsc]

Sure, here's a small test case:

import pybullet as pb

pb.connect(pb.GUI)
vis = pb.createVisualShape(shapeType = pb.GEOM_SPHERE, radius = 1.0)
col = pb.createCollisionShape(shapeType = pb.GEOM_SPHERE, radius = 1.0)

def create_3_bodies():
    positions = [[ 0.0, 0.0, 0.0],
                 [ 1.0, 0.0, 0.0],
                 [ 2.0, 0.0, 0.0]]
    return pb.createMultiBody(baseMass = 1.0,
                              baseCollisionShapeIndex = col,
                              baseVisualShapeIndex = vis,
                              batchPositions = positions,
                              basePosition = [0, 0, 0],
                              baseOrientation = pb.getQuaternionFromEuler([0, 0, 0]))

bodies = []
# create 3 bodies
bodies += create_3_bodies()
print(bodies)
# [0, 1, 2]

# remove first body
pb.removeBody(bodies[0])
bodies.pop(0)
print(bodies)
# [1, 2]

# create additional bodies
bodies += create_3_bodies()
print(bodies)
# [1, 2, 2, 3, 4]
#     ^--^--repeated IDS!

[erwincoumans]

Thanks for the repro case, it is a bug in batch creation indeed, will have a look later.