@Ghostvv
I'd like to remove longest matching rule, rules shouldn't overlap
Current behavior
In case multiple rules match the current state, the RulePolicy will use the longest matching rule.
Desired behavior
In case rules overlap the RulePolicy should
The RulePolicy should be trained before the other policies to be able to fail early.
What does "overlap" mean
action_overlap after two actions action_1, action_2 while rule b predicts action_overlap after a preceding action action_2I think we even emit hard error, and stop training until rules are fixed. Therefore, RulePolicy should be always trained first
I see where you're going, but on the contrary we are currently not doing the same for conflicting stories. Actually we currently not even warn the user about conflicting stories 馃檲
without RulePolicy, it was tricky to do
updated the PR description 馃憤
Also train the RulePolicy first to fail early
@Ghostvv Should I handle this or you?
I can handle it as part of RulePolicy refactor