mealy-decompose

Tools to investigate decomposition of Mealy machines, aka finite state machines (FSMs). Notable entry points are:

• app/Main.hs is a heuristic to decompose finite state machines into multiple components, based on their outputs.

• other/decompose_fsm_optimise.py does the same, but optimally and with a SAT solver. This can only handle state spaces of at most a few hundred states.

• app/RandomGen.hs for generating FSMs which are decomposable.

How to run

The the haskell tools (tested with ghc 9.2.8 and ghc 9.10.1):

cabal run mealy-decompose -- <path-to-fsm>

For the python tools (tested with python 3.12):

pip3 install python-sat
python3 decompose_fsm_optimise.py -h

Notes on copar

In the first versions of this tool, I used the copar library: CoPaR (The Coalgebraic Partition Refiner). This is a great library and in particular very fast. Despite that, I switched to my own partition refinement implementations because of the reasons below. The last commit to use copar is 2b8b79a4.

• I needed witnesses for some tasks. So I not only construct a partition, but also a splitting tree, where are track all the actions needed to separate states.

• I prefer to use the actual strings (or other data type) for the states, inputs and outputs as opposed to Ints. This makes it easier to debug, and less error-prone, as one cannot mix these entities anymore.

• copar comes with many dependencies, and it was a bit tricky to get the versioning correct.

Copyright

(c) 2024 Joshua Moerman, Open Universiteit