1 Oct
2015
1 Oct
'15
2:02 a.m.
Hi Alexandre,
thank you very much for your comments.
I also received the suggestion to use aliases to avoid the long
[0 & !1 & !2 & !3 & ...]
expressions in the body of the file.
Now, the output of our library looks like in the attachment.
According to ltl2tgba the output is correct (... for the automata/words that I
checked).
Best,
Matthias
On Tuesday, September 29, 2015 18:00:57 Alexandre Duret-Lutz wrote:
On Tue, Sep 29, 2015 at 5:09 PM, Matthias Heizmann
<heizmann(a)informatik.uni-freiburg.de> wrote:
Hi Spot developers,
I just extended our automata library
https://ultimate.informatik.uni-freiburg.de/?ui=int&tool=automata_scrip…
nterpreter&task=automataScript&sample=1954831617 with support to write
Büchi automata in the HOA format.
Sweet!
Now, I just wanted to check a few results without
implementing
support to read HOA files.
Syntactically my files seem to be correct, (example attached, file
is read by Spot), but I also wanted to do some semantic check,
e.g., test if a certain word is accepted.
From a semantic point of view, I suspect that you may have a problem
because the HOA format assumes that the alphabet is 2^AP (i.e., each
letter is an assignment of all atomic propositions), not AP.
So the automaton you attached also accepts an infinite word where each
letter is a&b, which somehow contradicts the name of the file.
Spot provides quite a number of sophisticated
operations but I did not
find something "simple" like an acceptance check.
Have I overlooked it or is it not documented?
For my purpose it is sufficient if it works on the "usual" Büchi automata.
You did not overlook anything: Spot does not export such a feature.
I'll add that to the todo list.
Currently, you can still check acceptance of a word by representing
this word as an automaton, and testing the interstection of the two
automata using autfilt's --intersect option.
If you can write your word as an LTL or PSL formula, is it even
easier than writing an automaton.
E.g. testing the word a&!b ; b&!a ; followed by a cycle of just a&b
% ltl2tgba -f "(a & !b) & X(b & !a) & XXG(a & b)" -H |
autfilt -q --intersect FinitelyManyAWithSink.hoa - &&
echo accepted || echo non-accepted
accepted
I guess you could restrict the FinitelyManyAWithSink.hoa
to disallow a and b to be accepted at the same time.
autfilt --exclusive-ap can be handy for that:
% autfilt --exclusive-ap=a,b FinitelyManyAWithSink.hoa -H > F2.hoa
Now you can evaluate words (via LTL) as if the alphabet was AP:
% ltl2tgba -f "a & X(b) & XXG(b)" -H |
autfilt -q --intersect F2.hoa - && echo accepted || echo non-accepted
accepted
% ltl2tgba -f "a & X(b) & XXG(a)" -H |
autfilt -q --intersect F2.hoa - && echo accepted || echo non-accepted
non-accepted
(Note that --intersect is a filter. Without the -q option it returns the
automata that interesect the reference automaton. If you want to
compute the product of two automata to check the intersection
by yourself, use --product instead.)
HTH