While turning an experiment public, few dataformats are listed in the sharing preferences, while there are public. I think this is not expected.

More specifically, these dataformats are the ones used by a private algorithm, whose sharing preferences are going to change, when the experiment become public.

In addition, I've noticed an inconsistency in the following two methods of the Experiment class in beat/web/experiments/models.py

• __share_dataformats line 211: errors.extend(needed_format.is_accessible(users=users, teams=teams))

• __share_algorithms line 238: errors.extend(needed_algorithm.is_accessible(False))

The method is_accessible() being common to both classes, the False argument is surprising. It is defined in beat/web/commons/models.py 155 def is_accessible(self, users=None, teams=None):

To reproduce the problem.

1. Create a new (private) version of an existing public algorithm
2. Run an algorithm with this private version of the algorithm
3. Try to make the experiment public