bug fix and improvement

[saeedark]

1. adds counter roll for the buffer_roll method
2. fixes no execution list when the network object itself has behavior
3. more efficient event recorder and support for multidimensional data
4. fix the remove_behavior bug for not passing the parent object.

[gitmv]

Hello, thanks for you contribution.
The first three updates seem to work and i will merge them, but the fourth one crashes because the arguments are switched.

Can i ask why you want to change the _remove_behavior_from_sorted_execution_list function in the first place, so that it works with no parent object?

The function is only used internally by the remove_behavior function in the NetworkObjectBase class. if you execute it without the parent, it can also remove multiple behaviours(if they are the same instance) of different objects (unlikely but it is possible that someone attaches the same behaviour instance twice) from the sorted_behavior_execution_list. The problem here is, that this behaviour is not only stored in the execution list, but also in the object.behaviours list. If you call net._remove_behavior_from_sorted_execution_list(beh), it is removed from the execution list, and not called anymore during a simulation run, but it would still be tag searchable and it would still appear in the obj.behaviours list.

Therefore _remove_behavior_from_sorted_execution_list should not be called directly from outside and behaviours should always be removed via the obj.remove_behavior(beh) function so that it is removed completely.

Do you have some reason why you want to use it this way?

[saeedark]

Hello,
Sorry, I didn't think that through. my bad.

But _remove_behavior_from_sorted_execution_list always removes the last instance of a behavior in an object. In that unlikely situation(same instance, multiple keys) removing a behavior with its key (via remove_behavior ) may lead to a wrong order of execution.

[saeedark]

the last commit should make _remove_behavior_from_sorted_execution_list specific to remove exact behavior.

[gitmv]

Ok this looks good!
The merge seems to work.

I appreciate it and feel free to improve the project further if you have any ideas or if you find bugs of some kind. I also think that the version with the torch core is a great idea.