spot/src/tgbaalgos/gtec/gtec.hh

237 lines
8.7 KiB
C++

// Copyright (C) 2003, 2004, 2005 Laboratoire d'Informatique de Paris 6 (LIP6),
// département Systèmes Répartis Coopératifs (SRC), Université Pierre
// et Marie Curie.
//
// This file is part of Spot, a model checking library.
//
// Spot is free software; you can redistribute it and/or modify it
// under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// Spot is distributed in the hope that it will be useful, but WITHOUT
// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
// or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
// License for more details.
//
// You should have received a copy of the GNU General Public License
// along with Spot; see the file COPYING. If not, write to the Free
// Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
// 02111-1307, USA.
#ifndef SPOT_TGBAALGOS_GTEC_GTEC_HH
# define SPOT_TGBAALGOS_GTEC_GTEC_HH
#include <stack>
#include "status.hh"
#include "tgbaalgos/emptiness.hh"
#include "tgbaalgos/emptiness_stats.hh"
namespace spot
{
/// \addtogroup emptiness_check_algorithms
/// @{
/// \brief Check whether the language of an automate is empty.
///
/// This is based on the following paper.
/// \verbatim
/// @InProceedings{couvreur.99.fm,
/// author = {Jean-Michel Couvreur},
/// title = {On-the-fly Verification of Temporal Logic},
/// pages = {253--271},
/// editor = {Jeannette M. Wing and Jim Woodcock and Jim Davies},
/// booktitle = {Proceedings of the World Congress on Formal Methods in
/// the Development of Computing Systems (FM'99)},
/// publisher = {Springer-Verlag},
/// series = {Lecture Notes in Computer Science},
/// volume = {1708},
/// year = {1999},
/// address = {Toulouse, France},
/// month = {September},
/// isbn = {3-540-66587-0}
/// }
/// \endverbatim
///
/// A recursive definition of the algorithm would look as follows,
/// but the implementation is of course not recursive.
/// (<code>&lt;Sigma, Q, delta, q, F&gt;</code> is the automaton to
/// check, H is an associative array mapping each state to its
/// positive DFS order or 0 if it is dead, SCC is and ACC are two
/// stacks.)
///
/// \verbatim
/// check(<Sigma, Q, delta, q, F>, H, SCC, ACC)
/// if q is not in H // new state
/// H[q] = H.size + 1
/// SCC.push(<H[q], {}>)
/// forall <a, s> : <q, _, a, s> in delta
/// ACC.push(a)
/// res = check(<Sigma, Q, delta, s, F>, H, SCC, ACC)
/// if res
/// return res
/// <n, _> = SCC.top()
/// if n = H[q]
/// SCC.pop()
/// mark_reachable_states_as_dead(<Sigma, Q, delta, q, F>, H$)
/// return 0
/// else
/// if H[q] = 0 // dead state
/// ACC.pop()
/// return true
/// else // state in stack: merge SCC
/// all = {}
/// do
/// <n, a> = SCC.pop()
/// all = all union a union { ACC.pop() }
/// until n <= H[q]
/// SCC.push(<n, all>)
/// if all != F
/// return 0
/// return new emptiness_check_result(necessary data)
/// \endverbatim
///
/// check() returns 0 iff the automaton's language is empty. It
/// returns an instance of emptiness_check_result. If the automaton
/// accept a word. (Use emptiness_check_result::accepting_run() to
/// extract an accepting run.)
///
/// There are two variants of this algorithm: spot::couvreur99_check and
/// spot::couvreur99_check_shy. They differ in their memory usage, the
/// number for successors computed before they are used and the way
/// the depth first search is directed.
///
/// spot::couvreur99_check performs a straightforward depth first search.
/// The DFS stacks store tgba_succ_iterators, so that only the
/// iterators which really are explored are computed.
///
/// spot::couvreur99_check_shy tries to explore successors which are
/// visited states first. this helps to merge SCCs and generally
/// helps to produce shorter counter-examples. However this
/// algorithm cannot stores unprocessed successors as
/// tgba_succ_iterators: it must compute all successors of a state
/// at once in order to decide which to explore first, and must keep
/// a list of all unexplored successors in its DFS stack.
///
/// The \c poprem parameter specifies how the algorithm should
/// handle the destruction of non-accepting maximal strongly
/// connected components. If \c poprem is true, the algorithm will
/// keep a list of all states of a SCC that are fully processed and
/// should be removed once the MSCC is popped. If \c poprem is
/// false, the MSCC will be traversed again (i.e. generating the
/// successors of the root recursively) for deletion. This is
/// a choice between memory and speed.
class couvreur99_check: public emptiness_check, public ec_statistics
{
public:
couvreur99_check(const tgba* a,
bool poprem = true,
const numbered_state_heap_factory* nshf
= numbered_state_heap_hash_map_factory::instance());
virtual ~couvreur99_check();
/// Check whether the automaton's language is empty.
virtual emptiness_check_result* check();
virtual std::ostream& print_stats(std::ostream& os) const;
/// \brief Return the status of the emptiness-check.
///
/// When check() succeed, the status should be passed along
/// to spot::counter_example.
///
/// This status should not be deleted, it is a pointer
/// to a member of this class that will be deleted when
/// the couvreur99 object is deleted.
const couvreur99_check_status* result() const;
protected:
couvreur99_check_status* ecs_;
/// \brief Remove a strongly component from the hash.
///
/// This function remove all accessible state from a given
/// state. In other words, it removes the strongly connected
/// component that contains this state.
void remove_component(const state* start_delete);
/// Whether to store the state to be removed.
bool poprem_;
};
/// \brief A version of spot::couvreur99_check that tries to visit
/// known states first.
///
/// If \a group is true (the default), the successors of all the
/// states that belong to the same SCC will be considered when
/// choosing a successor. Otherwise, only the successor of the
/// topmost state on the DFS stack are considered.
///
/// See the documentation for spot::couvreur99_check
class couvreur99_check_shy : public couvreur99_check
{
public:
couvreur99_check_shy(const tgba* a,
bool poprem = true,
bool group = true,
const numbered_state_heap_factory* nshf
= numbered_state_heap_hash_map_factory::instance());
virtual ~couvreur99_check_shy();
virtual emptiness_check_result* check();
protected:
struct successor {
bdd acc;
const spot::state* s;
successor(bdd acc, const spot::state* s): acc(acc), s(s) {}
};
// We use five main data in this algorithm:
// * couvreur99_check::root, a stack of strongly connected components (SCC),
// * couvreur99_check::h, a hash of all visited nodes, with their order,
// (it is called "Hash" in Couvreur's paper)
// * arc, a stack of acceptance conditions between each of these SCC,
std::stack<bdd> arc;
// * num, the number of visited nodes. Used to set the order of each
// visited node,
int num;
// * todo, the depth-first search stack. This holds pairs of the
// form (STATE, SUCCESSORS) where SUCCESSORS is a list of
// (ACCEPTANCE_CONDITIONS, STATE) pairs.
typedef std::list<successor> succ_queue;
struct todo_item
{
const state* s;
int n;
succ_queue q;
todo_item(const state* s, int n)
: s(s), n(n)
{
}
};
typedef std::list<todo_item> todo_list;
todo_list todo;
void clear_todo();
// Whether successors should be grouped for states in the same
// SCC.
bool group_;
/// \brief find the SCC number of a unprocessed state.
///
/// Sometimes we want to modify some of the above structures when
/// looking up a new state. This happens for instance when find()
/// must perform inclusion checking and add new states to process
/// to TODO during this step. (Because TODO must be known,
/// sub-classing spot::numbered_state_heap is not enough.) Then
/// overriding this method is the way to go.
virtual int* find_state(const state* s);
};
/// @}
}
#endif // SPOT_TGBAALGOS_GTEC_GTEC_HH