spot/doc/org/ltlgrind.org

# -*- coding: utf-8 -*-
#+TITLE: =ltlgrind=
#+DESCRIPTION: Spot command-line tool for mutating LTL formulas.
#+INCLUDE: setup.org
#+HTML_LINK_UP: tools.html
#+PROPERTY: header-args:sh :results verbatim :exports both

This tool lists formulas that are similar to but simpler than a given
formula by applying simple mutations to it, like removing operands or
formulas that are similar to but simpler than a given formula by
applying simple mutations to it, like removing operands or
operators. This is meant to be used with ltlcross to simplify a
formula that caused a problem before trying to debug it (see also
=ltlcross --grind=FILENAME=).

Here is a list of the different kinds of mutations that can be applied:

#+BEGIN_SRC sh :exports results
ltlgrind --help | sed -n '/Mutation rules.*:/,/^$/p' | sed '1d;$d'
#+END_SRC

#+RESULTS:
#+begin_example
      --ap-to-const          atomic propositions are replaced with true/false
      --remove-multop-operands   remove one operand from multops
      --remove-one-ap        all occurrences of an atomic proposition are
                             replaced with another atomic proposition used in
                             the formula
      --remove-ops           replace unary/binary operators with one of their
                             operands
      --rewrite-ops          rewrite operators that have a semantically simpler
                             form: a U b becomes a W b, etc.
      --simplify-bounds      on a bounded unary operator, decrement one of the
                             bounds, or set min to 0 or max to unbounded
      --split-ops            when an operator can be expressed as a
                             conjunction/disjunction using simpler operators,
                             each term of the conjunction/disjunction is a
                             mutation. e.g. a <-> b can be written as ((a & b)
                             | (!a & !b)) or as ((a -> b) & (b -> a)) so those
                             four terms can be a mutation of a <-> b
#+end_example

By default, all rules are applied. But if one or more rules are
specified, only those will be applied.

#+BEGIN_SRC sh
ltlgrind -f 'a U G(b <-> c)' --split-ops --rewrite-ops --remove-ops
#+END_SRC

#+RESULTS:
#+begin_example
a
G(b <-> c)
a W G(b <-> c)
a U (b <-> c)
a U Gb
a U Gc
a U G(b -> c)
a U G(c -> b)
a U G(b & c)
a U G(!b & !c)
#+end_example

The idea behind this tool is that when a bogus algorithm is found with
=ltlcross=, you probably want to debug it using a smaller formula than
the one found by =ltlcross=. So you would give the formula found by
=ltlcross= as an argument to =ltlgrind= and then use the resulting
mutations as a new input for =ltlcross=. It might report an error on
one of the mutation, which is guaranteed to be simpler than the
initial formula. The process can then be repeated until no error is
reported by =ltlcross=.

Since the whole process can become quite tedious, it can be done
automatically by calling =ltlcross= with the =--grind=FILENAME=
argument.

* Miscellaneous options
** =--sort=
  Output formulas from the shortest to the longest one. The
  length of a formula is the number of atomic propositions, constants
  and operators occurring in the formula.
** =-m N=, =--mutations=N=
  Specify the number of mutations to be applied to the formula. By
  default, =N=1=, so using this option is like calling ltlgrind on its
  own output several times, except for the fact that, when called on
  several formulas, ltlgrind doesn't handle duplicates.
** =-n N=, =--max-output=N=
   Limit the number of mutated formulas output to =N=.

#  LocalWords:  ltlgrind num toc html ltlcross FILENAME SRC sed ap Gb
#  LocalWords:  const multop multops unary decrement disjunction Gc
#  LocalWords:  utf args