def test_neighbor_search():
np.random.seed(0x4d3d3d3)
ds = fake_particle_ds(npart=16**3)
ds.periodicity = (True, True, True)
ds.index
fn, = add_nearest_neighbor_field("all", "particle_position", ds)
dd = ds.all_data()
nearest_neighbors = dd[fn]
pos = dd["particle_position"]
all_neighbors = np.zeros_like(nearest_neighbors)
any_eq = np.zeros(pos.shape[0], dtype='bool')
min_in = np.zeros(pos.shape[0], dtype='int64')
for i in range(pos.shape[0]):
dd.set_field_parameter("center", pos[i, :])
#radius = dd["particle_radius"]
#radius.sort()
r2 = (pos[:, 0] * pos[:, 0]) * 0
for j in range(3):
DR = (pos[i, j] - pos[:, j])
DRo = DR.copy()
DR[DRo > ds.domain_width[j] / 2.0] -= ds.domain_width[j]
DR[DRo < -ds.domain_width[j] / 2.0] += ds.domain_width[j]
r2 += DR * DR
radius = np.sqrt(r2)
iii = np.argsort(radius)
radius.sort()
assert (radius[0] == 0.0)
all_neighbors[i] = radius[63]
any_eq[i] = np.any(np.abs(radius - nearest_neighbors[i]) < 1e-7)
min_in[i] = np.argmin(np.abs(radius - nearest_neighbors[i]))
#if i == 34: raise RuntimeError
#dd.field_data.pop(("all", "particle_radius"))
yield assert_equal, (min_in == 63).sum(), min_in.size
yield assert_array_almost_equal, nearest_neighbors, all_neighbors
def _smoothing_length(field, data):
# For now, we hardcode num_neighbors. We should make this configurable
# in the future.
num_neighbors = 64
fn, = add_nearest_neighbor_field(ptype, "particle_position", self,
num_neighbors)
return data[ptype, 'nearest_neighbor_distance_%d' % num_neighbors]
def test_neighbor_search():
np.random.seed(0x4d3d3d3)
ds = fake_particle_ds(npart = 16**3)
ds.periodicity = (True, True, True)
ds.index
fn, = add_nearest_neighbor_field("all", "particle_position", ds)
dd = ds.all_data()
nearest_neighbors = dd[fn]
pos = dd["particle_position"]
all_neighbors = np.zeros_like(nearest_neighbors)
any_eq = np.zeros(pos.shape[0], dtype='bool')
min_in = np.zeros(pos.shape[0], dtype='int64')
for i in range(pos.shape[0]):
dd.set_field_parameter("center", pos[i,:])
#radius = dd["particle_radius"]
#radius.sort()
r2 = (pos[:,0]*pos[:,0])*0
for j in range(3):
DR = (pos[i,j] - pos[:,j])
DRo = DR.copy()
DR[DRo > ds.domain_width[j]/2.0] -= ds.domain_width[j]
DR[DRo < -ds.domain_width[j]/2.0] += ds.domain_width[j]
r2 += DR*DR
radius = np.sqrt(r2)
iii = np.argsort(radius)
radius.sort()
assert(radius[0] == 0.0)
all_neighbors[i] = radius[63]
any_eq[i] = np.any( np.abs(radius - nearest_neighbors[i]) < 1e-7 )
min_in[i] = np.argmin(np.abs(radius - nearest_neighbors[i]))
#if i == 34: raise RuntimeError
#dd.field_data.pop(("all", "particle_radius"))
yield assert_equal, (min_in == 63).sum(), min_in.size
yield assert_array_almost_equal, nearest_neighbors, all_neighbors
def setup_gas_particle_fields(self, ptype):
num_neighbors = 65
fn, = add_nearest_neighbor_field(ptype, "particle_position", self, num_neighbors)
def _func():
def _smoothing_length(field, data):
# For now, we hardcode num_neighbors. We should make this configurable
# in the future.
rv = data[ptype, 'nearest_neighbor_distance_%d' % num_neighbors]
#np.maximum(rv, 0.5*data[ptype, "Epsilon"], rv)
return rv
return _smoothing_length
self.add_field(
(ptype, "smoothing_length"),
function=_func(),
particle_type=True,
units="code_length")
def test_neighbor_search():
# skip for now, in principle we can reimplement this in the demeshening
import nose
raise nose.SkipTest
np.random.seed(0x4D3D3D3)
ds = fake_particle_ds(npart=16 ** 3)
ds.periodicity = (True, True, True)
ds.index
(fn,) = add_nearest_neighbor_field("all", "particle_position", ds)
dd = ds.all_data()
nearest_neighbors = dd[fn]
pos = dd["particle_position"]
all_neighbors = np.zeros_like(nearest_neighbors)
any_eq = np.zeros(pos.shape[0], dtype="bool")
min_in = np.zeros(pos.shape[0], dtype="int64")
for i in range(pos.shape[0]):
dd.set_field_parameter("center", pos[i, :])
# radius = dd["particle_radius"]
# radius.sort()
r2 = (pos[:, 0] * pos[:, 0]) * 0
for j in range(3):
DR = pos[i, j] - pos[:, j]
DRo = DR.copy()
DR[DRo > ds.domain_width[j] / 2.0] -= ds.domain_width[j]
DR[DRo < -ds.domain_width[j] / 2.0] += ds.domain_width[j]
r2 += DR * DR
radius = np.sqrt(r2)
radius.sort()
assert radius[0] == 0.0
all_neighbors[i] = radius[63]
any_eq[i] = np.any(np.abs(radius - nearest_neighbors[i]) < 1e-7)
min_in[i] = np.argmin(np.abs(radius - nearest_neighbors[i]))
# if i == 34: raise RuntimeError
# dd.field_data.pop(("all", "particle_radius"))
assert_equal((min_in == 63).sum(), min_in.size)
assert_array_almost_equal(nearest_neighbors, all_neighbors)
def _smoothing_length(field, data):
# For now, we hardcode num_neighbors. We should make this configurable
# in the future.
num_neighbors = 64
fn, = add_nearest_neighbor_field(ptype, "particle_position", self, num_neighbors)
return data[ptype, 'nearest_neighbor_distance_%d' % num_neighbors]
