class TestSolving(TestCase):
'''
Tests basic solving and related functions.
'''
def setUp(self):
self.mcb = _MCB()
self.mit = _MCB()
self.ctl = Control(['0'])
def tearDown(self):
self.mcb = None
self.mit = None
self.ctl = None
def test_solve_result_str(self):
'''
Test string representation of solve results.
'''
ret = self.ctl.solve()
self.assertEqual(str(ret), 'SAT')
self.assertRegex(repr(ret), 'SolveResult(.*)')
def test_model_str(self):
'''
Test string representation of models.
'''
self.ctl.add('base', [], 'a.')
self.ctl.ground([('base', [])])
with self.ctl.solve(yield_=True) as hnd:
for mdl in hnd:
self.assertEqual(str(mdl), "a")
self.assertRegex(repr(mdl), "Model(.*)")
def test_solve_cb(self):
'''
Test solving using callback.
'''
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
_check_sat(self, cast(SolveResult, self.ctl.solve(on_model=self.mcb.on_model, yield_=False, async_=False)))
self.assertEqual(self.mcb.models, _p(['a', 'c'], ['b', 'c']))
self.assertEqual(self.mcb.last[0], ModelType.StableModel)
def test_solve_async(self):
'''
Test asynchonous solving.
'''
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
with cast(SolveHandle, self.ctl.solve(on_model=self.mcb.on_model, yield_=False, async_=True)) as hnd:
_check_sat(self, hnd.get())
self.assertEqual(self.mcb.models, _p(['a', 'c'], ['b', 'c']))
def test_solve_yield(self):
'''
Test solving yielding models.
'''
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
with cast(SolveHandle, self.ctl.solve(on_model=self.mcb.on_model, yield_=True, async_=False)) as hnd:
for m in hnd:
self.mit.on_model(m)
_check_sat(self, hnd.get())
self.assertEqual(self.mcb.models, _p(['a', 'c'], ['b', 'c']))
self.assertEqual(self.mit.models, _p(['a', 'c'], ['b', 'c']))
def test_solve_async_yield(self):
'''
Test solving yielding models asynchronously.
'''
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
with self.ctl.solve(on_model=self.mcb.on_model, yield_=True, async_=True) as hnd:
while True:
hnd.resume()
_ = hnd.wait()
m = hnd.model()
if m is None:
break
self.mit.on_model(m)
_check_sat(self, hnd.get())
self.assertEqual(self.mcb.models, _p(['a', 'c'], ['b', 'c']))
self.assertEqual(self.mit.models, _p(['a', 'c'], ['b', 'c']))
def test_solve_interrupt(self):
'''
Test interrupting solving.
'''
self.ctl.add("base", [], "1 { p(P,H): H=1..99 } 1 :- P=1..100.\n1 { p(P,H): P=1..100 } 1 :- H=1..99.")
self.ctl.ground([("base", [])])
with self.ctl.solve(async_=True) as hnd:
hnd.resume()
hnd.cancel()
ret = hnd.get()
self.assertTrue(ret.interrupted)
with self.ctl.solve(async_=True) as hnd:
hnd.resume()
self.ctl.interrupt()
ret = hnd.get()
self.assertTrue(ret.interrupted)
def test_solve_core(self):
'''
Test core retrieval.
'''
self.ctl.add("base", [], "3 { p(1..10) } 3.")
self.ctl.ground([("base", [])])
ass = []
for atom in self.ctl.symbolic_atoms.by_signature("p", 1):
ass.append(-atom.literal)
ret = cast(SolveResult, self.ctl.solve(on_core=self.mcb.on_core, assumptions=ass))
self.assertTrue(ret.unsatisfiable)
self.assertTrue(len(self.mcb.core) > 7)
def test_enum(self):
'''
Test core retrieval.
'''
self.ctl = Control(['0', '-e', 'cautious'])
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
self.ctl.solve(on_model=self.mcb.on_model)
self.assertEqual(self.mcb.last[0], ModelType.CautiousConsequences)
self.assertEqual([self.mcb.last[1]], _p(['c']))
self.ctl = Control(['0', '-e', 'brave'])
self.ctl.add("base", [], "1 {a; b} 1. c.")
self.ctl.ground([("base", [])])
self.ctl.solve(on_model=self.mcb.on_model)
self.assertEqual(self.mcb.last[0], ModelType.BraveConsequences)
self.assertEqual([self.mcb.last[1]], _p(['a', 'b', 'c']))
def test_model(self):
'''
Test functions of model.
'''
def on_model(m: Model):
self.assertTrue(m.contains(Function('a')))
self.assertTrue(m.is_true(cast(SymbolicAtom, m.context.symbolic_atoms[Function('a')]).literal))
self.assertFalse(m.is_true(1000))
self.assertEqual(m.thread_id, 0)
self.assertEqual(m.number, 1)
self.assertFalse(m.optimality_proven)
self.assertEqual(m.cost, [3])
m.extend([Function('e')])
self.assertSequenceEqual(m.symbols(theory=True), [Function('e')])
self.ctl.add("base", [], "a. b. c. #minimize { 1,a:a; 1,b:b; 1,c:c }.")
self.ctl.ground([("base", [])])
self.ctl.solve(on_model=on_model)
def test_control_clause(self):
'''
Test adding clauses while solving.
'''
self.ctl.add("base", [], "1 {a; b; c} 1.")
self.ctl.ground([("base", [])])
with cast(SolveHandle, self.ctl.solve(on_model=self.mcb.on_model, yield_=True, async_=False)) as hnd:
for m in hnd:
clause = []
if m.contains(Function('a')):
clause.append((Function('b'), False))
else:
clause.append((Function('a'), False))
m.context.add_clause(clause)
_check_sat(self, hnd.get())
self.assertEqual(len(self.mcb.models), 2)
def test_control_nogood(self):
'''
Test adding nogoods while solving.
'''
self.ctl.add("base", [], "1 {a; b; c} 1.")
self.ctl.ground([("base", [])])
with cast(SolveHandle, self.ctl.solve(on_model=self.mcb.on_model, yield_=True, async_=False)) as hnd:
for m in hnd:
clause = []
if m.contains(Function('a')):
clause.append((Function('b'), True))
else:
clause.append((Function('a'), True))
m.context.add_nogood(clause)
_check_sat(self, hnd.get())
self.assertEqual(len(self.mcb.models), 2)
class VizloControl(Control):
def add_to_painter(self, model: Union[Model, PythonModel,
Collection[clingo.Symbol]]):
"""
will register model with the internal painter. On all consecutive calls to paint(), this model will be painted.
:param model: the model to add to the painter.
:return:
"""
self.painter.append(PythonModel(model))
def __init__(self,
arguments: List[str] = [],
logger=None,
message_limit: int = 20,
print_entire_models=False,
atom_draw_maximum=15):
self.control = Control(arguments, logger, message_limit)
self.painter: List[PythonModel] = list()
self.program: ASTProgram = list()
self.raw_program: str = ""
self.transformer = JustTheRulesTransformer()
self._print_changes_only = not print_entire_models
self._atom_draw_maximum = atom_draw_maximum
def _set_print_only_changes(self, value: bool) -> None:
self._print_changes_only = value
def ground(self,
parts: List[Tuple[str, List[Symbol]]],
context: Any = None) -> None:
self.control.ground(parts, context)
def solve(self,
assumptions: List[Union[Tuple[Symbol, bool], int]] = [],
on_model=None,
on_statistics=None,
on_finish=None,
yield_: bool = False,
async_: bool = False) -> Union[SolveHandle, SolveResult]:
return self.control.solve(assumptions, on_model, on_statistics,
on_finish, yield_, async_)
def load(self, path):
prg = ""
with open(path) as f:
for line in f:
prg += line
self.program += prg
self.control.load(path)
def add(self, name: str, parameters: List[str], program: str) -> None:
self.raw_program += program
self.control.add(name, parameters, program)
def find_nodes_corresponding_to_stable_models(self, g, stable_models):
correspoding_nodes = set()
for model in stable_models:
for node in g.nodes():
log(f"{node} {type(node.model)} == {model} {type(model)} -> {set(node.model) == model}"
)
if set(node.model) == model and len(
g.edges(node)) == 0: # This is a leaf
log(f"{node} <-> {model}")
correspoding_nodes.add(node)
break
return correspoding_nodes
def prune_graph_leading_to_models(self, graph: nx.DiGraph,
models_as_nodes):
before = len(graph)
relevant_nodes = set()
for model in models_as_nodes:
for relevant_node in nx.all_simple_paths(graph, INITIAL_EMPTY_SET,
model):
relevant_nodes.update(relevant_node)
all_nodes = set(graph.nodes())
irrelevant_nodes = all_nodes - relevant_nodes
graph.remove_nodes_from(irrelevant_nodes)
after = len(graph)
log(f"Removed {before - after} of {before} nodes ({(before - after) / before})"
)
def _make_graph(self, _sort=True):
"""
Ties together transformation and solving. Transforms the already added program parts and creates a solving tree.
:param _sort: Whether the program should be sorted automatically. Setting this to false will likely result into
wrong results!
:return:
:raises ValueError:
"""
if not len(self.raw_program):
raise ValueError("Can't paint an empty program.")
else:
t = JustTheRulesTransformer()
program = t.transform(self.raw_program, _sort)
if len(self.painter):
universe = get_ground_universe(program)
global_assumptions = make_global_assumptions(
universe, self.painter)
solve_runner = SolveRunner(program, t.rule2signatures)
g = solve_runner.make_graph(global_assumptions)
else:
solve_runner = SolveRunner(program,
symbols_in_heads_map=t.rule2signatures)
g = solve_runner.make_graph()
return g
def paint(self,
atom_draw_maximum: int = 20,
show_entire_model: bool = False,
sort_program: bool = True,
**kwargs):
"""
Will create a graph visualization of the solving process. If models have been added using add_to_painter,
only the solving paths that lead to these models will be drawn.
:param atom_draw_maximum: int
The maximum amount of atoms that will be printed for each partial model. (default=20)
:param show_entire_model: bool
If false, only the atoms that have been added at a solving step will be printed (up to atom_draw_maximum).
If true, all atoms will always be printed (up to atom_draw_maximum). (default=False)
:param sort_program:
If true, the rules of a program will be sorted and grouped by their dependencies.
Each set of rules will contain all rules in which each atom in its heads is contained in a head.
:param kwargs:
kwargs will be forwarded to the visualisation module. See graph.draw()
:return:
"""
if type(atom_draw_maximum) != int:
raise ValueError(
f"Argument atom_draw_maximum should be an integer (received {atom_draw_maximum})."
)
g = self._make_graph(sort_program)
display = NetworkxDisplay(g, atom_draw_maximum, not show_entire_model)
img = display.draw(**kwargs)
return img
def _add_and_ground(self, prg):
"""Short cut for complex add and ground calls, should only be used for debugging purposes."""
self.add("base", [], prg)
self.ground([("base", [])])
##################
# Just pass-through stuff
##################
@property
def configuration(self) -> Configuration:
return self.control.configuration
@property
def is_conflicting(self) -> bool:
return self.control.is_conflicting
@property
def statistics(self) -> dict:
return self.control.statistics
@property
def symbolic_atoms(self) -> SymbolicAtoms:
return self.control.symbolic_atoms
@property
def theory_atoms(self) -> TheoryAtomIter:
return self.control.theory_atoms
@property
def use_enumeration_assumption(self) -> bool:
return self.control.use_enumeration_assumption
def assign_external(self, external: Union[Symbol, int],
truth: Optional[bool], **kwargs) -> None:
self.control.assign_external(external, truth, **kwargs)
def backend(self) -> Backend:
return self.control.backend()
def builder(self) -> ProgramBuilder:
return self.control.builder()
def cleanup(self) -> None:
self.control.cleanup()
def get_const(self, name: str) -> Optional[Symbol]:
return self.control.get_const(name)
def interrupt(self):
self.control.interrupt()
def register_observer(self, observer, replace=False):
self.register_observer(observer, replace)
def release_external(self, symbol: Union[Symbol, int]) -> None:
self.control.release_external(symbol)