Open mdickinson opened 10 years ago
Maybe sort by weight and class depth (in the sense of the shortest path from the class to object
in the 'is_a_direct_subclass' DAG)?
See #883 for one approach to solving this; I've closed that PR, but we may want to pick up the code from that PR again as a basis for a solution at some point in the future (though I think there are also other approaches).
In
AdaptationManager._adapt
there's a line:But the
_by_weight_then_from_protocol_specificity
function doesn't give a transitive relation, so the results of the sort are undefined. It's not clear to me how important the sort order is here.