Ideals and sheep

[pazz]

when trying to write the csv parser I had a dive through the data structures used here and one thing was surprising/confusing:

What you called Ideal here is actually just a dw-closed set, and as such can be represented as finite union of incomparable ideals, which are called Sheep here.

cf def 4.6 In http://www.lsv.fr/~phs/algorithmic_aspects_of_wqos.pdf.

Another related thing that I'm confused about is how are you sure that the other of states in the Sheep vector. corresponds to the order of state names in the nfa after sorting? It looks as if you rely on calling iter() on a Hashset, which could traverse in arbitrary orders?
Thanks

[Numero7]

@pazz you are right this is confusing.
I did create the branch

https://github.com/SynthesisLab/shepherd/tree/69-ideals-and-sheep

to cleanup this mess.

What do you think?

About your other question, what occurence of iter() are you concerned with? The internal representation of states is u8 not strings. The reordering of states is performed by nfa::apply_reordering with some care for not introducing discrpeancies. And then the solver only manipulates nfa.get_edges() nfa.get_alphabet() nfa.final_states() nfa.initial_states() all dealing with the u8 representation

[pazz]

Thanks, I see now. I overlooked that sheep are created only after you resorted the states in the nfa.