Declare F as a list, not a set. Fixes #48.

[adl]

Here is a proposal for #48. The list idea is from Joachim.

The inversion of F and R in the description is because m, used in R, comes from the size of F.

[strejcek]

The suggested order of F and R is not lucky: R precedes F in the automaton tuple and definition of F also heavily uses R, so R should be defined first.

However, the question is whether we really want/need to define F: it seems that in the current setting it just duplicates the information already described by R. Wouldn't it be cleaner to define S_i sets just before the definition of an accepting run?

[adl]

On Wed, Jun 3, 2015 at 2:13 PM, Jan Strejček notifications@github.com wrote:

However, the question is whether we really want/need to define F: it seems that in the current setting it just duplicates the information already described by R. Wouldn't it be cleaner to define S_i sets just before the definition of an accepting run?

The reason we define the S_i set as part of the automaton is to make the parallel with state-based acceptance very clear. If you remove the F from the transition-based definition, you need a diferent definition for automata with state-based acceptance. (I'm not against it, I'm just explaining why things look the way they look.)

Alexandre Duret-Lutz

[strejcek]

Ah, I see. I'm not sure whether the benefits for state-based acceptance pays the unsightliness of the basic definition. Others?

[kleinj]

I agree that the S_i sets in the tuple for the transition-based automaton are redundant and would propose something in the spirit of the following:

• remove $F$ from the tuple for the transition-based automaton, replace by $m$ ($m$ is the number of acceptance sets used by the automaton), with the explanation for $R$ coming after $m$
• Afterwards, have The indizes of the acceptance sets appearing in the transitions induce a list of acceptance sets $F=(S_0, ..., S_{m-1})$ with $S_i \subseteq R$ containing those transitions with $i \in M$, i.e., ... [as currently]
• For state-based acceptance, use $F$ in the tuple and in the explanation for F have a note such as In contrast to the automaton with transition-based acceptance, the acceptance sets $S_0, ..., S_{m-1}$ are directly defined, with each S_i being a subset of states.

I think that this would actually improve the clarity in the transition-based case, and would not complicate the state-based definition.

[adl]

Superseded by #50.