How do I diagram an existing Machine for debugging?

[leingang]

I have a script which runs a machine and I wanted to check if the machine was configured the way I expected. I thought using the diagrams extension would allow me to generate a diagram for inspection.

But all the use cases in the documentation show how to create an object of class GraphMachine, add states and transitions to that object, and diagram it. If I have an existing Machine, how can I create a GraphMachine with the same states and transitions, and diagram that?

[aleneum]

Hello @leingang,

If I have an existing Machine, how can I create a GraphMachine with the same states and transitions, and diagram that?

unfortunately, transitions does not feature a copy constructor or method (yet). HierarchicalMachine does some shallow copying but expects the copied/embedded machine to be of the same class. GraphMachine and Machine are mostly the same. You could get a diagram from a temporary GraphMachine by shallow copying/assigning the relevant properties of your machine to the graph machine:

from transitions.extensions import GraphMachine
from transitions import Machine

m = Machine(states=["A", "B", "C", "D"], initial="C")
m.add_ordered_transitions(trigger="next", conditions=["is_A", "is_B", "is_C", "is_D"])
# do not initialise a model and replace events and states
tmp = GraphMachine(model=None, initial=None, show_conditions=True)
tmp.states = m.states
tmp.events = m.events
# we can pass machine as a model to our graph machine
tmp._get_graph(m).draw("graph_simple.png", prog="dot")

Alternatively, you could use MarkupMachine as your machine class. It features the generation of markup configurations that GraphMachine understands:

from transitions.extensions import GraphMachine
from transitions.extensions.markup import MarkupMachine as Machine

mm = Machine(states=["A", "B", "C", "D"], initial="A")
mm.add_ordered_transitions(trigger="next", conditions=["is_A", "is_B", "is_C", "is_D"])
mm.next()
gm = GraphMachine(markup=mm.markup, show_conditions=True)
gm.next()
gm.get_graph().draw("graph_markup.png", prog="dot")

This will create a new copy of all states and transitions. Thus, gm and mm can be used independently. But if it is just about debugging I'd suggest to alter your import statement:

if DEBUG:
    from transitions.extensions import GraphMachine as Machine
else:
    from transitions import Machine

[...]
if DEBUG:
    machine.get_graph()...

Last but not least: Many people do not add states and transitions via add_transitions or add_states but only use Python dictionaries or YAML/JSON to configure state machines. So, if you do not need to add transitions and/or states dynamically, this could also be a solution:

states = ["A", "B", "C", "D"]
transitions = [
    ["go", "A", "B"],
    ["check", "B", "C"],
    {"trigger": "undo", "source": "C", "dest": "D", "conditions": "has_failed"}
]
# ...
m = Machine(states=states, transitions=transitions, initial="A")
# ...
gm = GraphMachine(states=states, transitions=transitions, initial=m.state)

[leingang]

Thank you! I managed to solve my problem without diagramming the machine, but I'm still interested in doing this. So I will give your suggestions a shot.