Failed to compose DFAs about dfa HOT 6 CLOSED

hi, what version of python and the dfa package are you using?

from dfa.

Hi @tomyaacov

Sorry about the confusion. I just realized that there was a regression in the readme. The composition API was dropped a while back and it was removed from the readme.

It looks like it was accidentally brought back when cherry picking and old patch.

from dfa.

Ok.. thanks for replying. Is there an older version of the package which includes composition?

from dfa.

So it was explored in an experimental feature branch.

I actually think it might be good to try and explore adding the feature again, but I won't have time to really investigate (and then commit to maintaining) until ~May.

I'm happy to accept a PR though.

Are you just interested in parallel composition? Or do you generally want parallel + sequential composition?

For small alphabets, it's fairly easy to write code that implements these. For example parallel composition should be something like:

from dfa import DFA

def par_compose(dfas: list[DFA]) -> DFA:
    def transition(state, char):
        return tuple(d._transition(s, c) for d, s, c in zip(dfas, state, char))
        
   def label(state):
       return tuple(d._label(s) for s in zip(dfas, state))
       
   return DFA(
       start=tuple(d.start for d in dfas),
       inputs=product(d.inputs for d in dfas),
       outputs=product(d.outputs for d in dfas),
       label=label,
       transition=transition,
   )

from dfa.

To be clear, the real pain is writing logic to symbolically handle sum and product alphabets.

It's not impossible, but I recall there being subtle edge cases.

Typically, I do composition at the circuit level using the py-aiger library.

from dfa.

Will check it out. Thanks!

from dfa.