In AALpy helper functions that work on automata (e.g.make_input_complete) we always need to check which kind of automaton we are handling. Usually there is a big fat if-then-else thingy that has options for each automaton type and often those code paths are only needed to account for the miniature differences between automaton state (data) structures. For a while now I've been thinking that it would be nice to have a common interface for automata to get rid of this and make future development easier.
One thing that we could do is to implement a method for each automaton type that exposes a dynamic view of the automaton as an LTS with a common structure for labels for states and transitions. Transition labels could be behave like a dictionary with different keys for each automaton type. Moore machine transitions would have a single key"input". Transitions of stochastic Mealy machines would have"output" and"probability" in addition to that.
Take for example the part ofmake_input_complete that adds a new transition after having identified some states where inputi is missing. Currently, this has 24 lines of ifs. Using the proposed LTS view, we could write code similar to this:

target_state=sifmissing_transition_go_to=='self_loop'elsesink_statelabel=dict(input=i)ifisinstance(automaton, (MealyMachine,Onfsm,StochasticMealyMachine)):label["output"]="epsilon"ifisinstance(automaton, (Mdp,StochasticMealyMachine)):label["probability"]=1.new_transition=lts_view.add_transition(source=s,target=target_state,label=label)

Ideally minus the magic strings :D

If we use mixins for the distinctions moore vs mealy and det vs non-det vs stochastic the comparison gets even simpler and more robust. Then we could replace e.g.isinstance(automaton, (MealyMachine, Onfsm, StochasticMealyMachine)) by - let's say -isinstance(automaton, MealyMachineLike).

Let me know what you think.