Is GeometryObject::ngeoms really useful ?

[florent-lamiraux]

https://github.com/stack-of-tasks/pinocchio/blob/593d4d43fded997bb9aa2421f4e55294dbd233c4/src/multibody/geometry.hpp#L168 Is there any difference between ngeoms and geometryObjects ? If no, the first one should be deprecated and replaced by a const reference to the second one. For some robotics demo, I add and remove geometries from my model and I got confused since geometryObjects.erase(it) needs to be coupled with ngeoms--.

By the way, I wrote a function that removes a GeometryObject instance from a GeometryModel and removes the corresponding collision pairs. The input of the function is the name of the object. Do you think this function should be added to pinocchio ?

[jcarpent]

By the way, I wrote a function that removes a GeometryObject instance from a GeometryModel and removes the corresponding collision pairs. The input of the function is the name of the object. Do you think this function should be added to pinocchio ?

Yes, of course.

[jcarpent]

Is there any difference between ngeoms and geometryObjects ? If no, the first one should be deprecated and replaced by a const reference to the second one. For some robotics demo, I add and remove geometries from my model and I got confused since geometryObjects.erase(it) needs to be coupled with ngeoms--.

Why not. How do you see the solution?

[florent-lamiraux]

In class GeometryModel:

const Index& ngeoms;

and in the constructor:

    GeometryModel()
    : ngeoms(geometryObjects.size())
    , geometryObjects()
    , collisionPairs()
    {}

If the compiler accepts, this should do the job.

[jcarpent]

It will lead to some memory leaks I guess, as std::vector<T>::size follows this signature:

size_type size() const noexcept;

[florent-lamiraux]

It will lead to some memory leaks I guess, as std::vector<T>::size follows this signature:

I do not think so. There is no memory allocation involved.

[jcarpent]

Yes, but the return object is a temporary variable: it is not held by the std::vector object.

[florent-lamiraux]

Let see what the compiler thinks about it. I will try tomorrow.

[jcarpent]

Let's then bet a 🍺 dear Florent!?

[florent-lamiraux]

I think you are right. My solution won't work. Then I will propose a PR to add a method that removes an object. Calling this method, users will not need to handle variable ngeoms.