spot/doc/org/tut10.org
Alexandre Duret-Lutz 22f442f758 parsetl: change the interface to return a parsed_formula
This gets the interface of all the functions parsing formula in line
with the interface of the automaton parser: both return a "parsed_*"
object (parsed_formula or parsed_automaton) that contains the said
object and its list of errors.  Doing so avoid having to declare the
parse_error_list in advance.

* spot/tl/parse.hh, spot/parsetl/parsetl.yy: Do the change.
* spot/parsetl/fmterror.cc: Adjust the error printer.
* NEWS: Document it.
* bin/common_finput.cc, bin/common_finput.hh, bin/ltlcross.cc,
bin/ltldo.cc, bin/ltlfilt.cc, doc/org/tut01.org, doc/org/tut02.org,
doc/org/tut10.org, doc/org/tut20.org, python/ajax/spotcgi.in,
python/spot/impl.i, spot/parseaut/parseaut.yy, tests/core/checkpsl.cc,
tests/core/checkta.cc, tests/core/consterm.cc, tests/core/emptchk.cc,
tests/core/equalsf.cc, tests/core/ikwiad.cc, tests/core/kind.cc,
tests/core/length.cc, tests/core/ltlprod.cc, tests/core/ltlrel.cc,
tests/core/randtgba.cc, tests/core/readltl.cc, tests/core/reduc.cc,
tests/core/safra.cc, tests/core/syntimpl.cc, tests/core/tostring.cc,
tests/ltsmin/modelcheck.cc, tests/python/alarm.py,
tests/python/interdep.py, tests/python/ltl2tgba.py,
tests/python/ltlparse.py: Adjust all uses.
2016-02-17 20:31:58 +01:00

4.6 KiB

Translating an LTL formula into a never claim

Here is how to translate an LTL (or PSL) formula into a never claim.

Shell

ltl2tgba --spin 'GFa -> GFb'
never { /* F(GFb | G!a) */
T0_init:
  if
  :: ((!(a))) -> goto accept_S0
  :: ((true)) -> goto T0_init
  :: ((b)) -> goto accept_S2
  fi;
accept_S0:
  if
  :: ((!(a))) -> goto accept_S0
  fi;
accept_S2:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
T0_S3:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
}

Python

The formula function returns a formula object (or raises a parse-error exception). Formula objects have a translate() method that returns an automaton, and the automata objects have a to_str method that can output in one of the supported syntaxes.

So the translation is actually a one-liner in Python:

import spot
print(spot.formula('GFa -> GFb').translate('BA').to_str('spin'))
never {
T0_init:
  if
  :: ((!(a))) -> goto accept_S0
  :: ((true)) -> goto T0_init
  :: ((b)) -> goto accept_S2
  fi;
accept_S0:
  if
  :: ((!(a))) -> goto accept_S0
  fi;
accept_S2:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
T0_S3:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
}

The above line can actually be made a bit shorter, because translate() can also be used as a function (as opposed to a method) that takes a formula (possibly as a string) as first argument:

import spot
print(spot.translate('GFa -> GFb', 'BA').to_str('spin'))
never {
T0_init:
  if
  :: ((!(a))) -> goto accept_S0
  :: ((true)) -> goto T0_init
  :: ((b)) -> goto accept_S2
  fi;
accept_S0:
  if
  :: ((!(a))) -> goto accept_S0
  fi;
accept_S2:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
T0_S3:
  if
  :: ((b)) -> goto accept_S2
  :: ((!(b))) -> goto T0_S3
  fi;
}

C++

All the translation pipeline (this include simplifying the formula, translating the simplified formula into an automaton, and simplifying the resulting automaton) is handled by the spot::translator object. This object can configured by calling set_type() to chose the type of automaton to output, set_level() to set the level of optimization (it's high by default), and set_pref() to set various preferences (like small or deterministic) or characteristic (complete, unambiguous) for the resulting automaton. Finally, the output as a never claim is done via the print_never_claim function.

  #include <string>
  #include <iostream>
  #include <spot/tl/parse.hh>
  #include <spot/tl/print.hh>
  #include <spot/twaalgos/translate.hh>
  #include <spot/twaalgos/neverclaim.hh>

  int main()
  {
    std::string input = "[]<>p0 || <>[]p1";
    spot::parsed_formula pf = spot::parse_infix_psl(input);
    if (pf.format_errors(std::cerr))
      return 1;
    spot::translator trans;
    trans.set_type(spot::postprocessor::BA);
    spot::twa_graph_ptr aut = trans.run(pf.f);
    print_never_claim(std::cout, aut) << '\n';
    return 0;
  }
never {
T0_init:
  if
  :: (p1) -> goto accept_S0
  :: (true) -> goto T0_init
  :: (p0) -> goto accept_S2
  fi;
accept_S0:
  if
  :: (p1) -> goto accept_S0
  fi;
accept_S2:
  if
  :: (p0) -> goto accept_S2
  :: (!(p0)) -> goto T0_S3
  fi;
T0_S3:
  if
  :: (p0) -> goto accept_S2
  :: (!(p0)) -> goto T0_S3
  fi;
}

Additional comments

The Python version of translate() is documented as follows:

import spot
help(spot.translate)
Help on function translate in module spot:

translate(formula, *args)
    Translate a formula into an automaton.

    Keep in mind that pref expresses just a preference that may not be
    satisfied.

    The optional arguments should be strings among the following:
    - at most one in 'TGBA', 'BA', or 'Monitor'
      (type of automaton to build)
    - at most one in 'Small', 'Deterministic', 'Any'
      (preferred characteristics of the produced automaton)
    - at most one in 'Low', 'Medium', 'High'
      (optimization level)
    - any combination of 'Complete', 'Unambiguous', and
      'StateBasedAcceptance' (or 'SBAcc' for short)

    The default correspond to 'tgba', 'small' and 'high'.