def time_tree_construction(Ntime, LStime, ndim=2):
pts, le, re, ls = make_points(Ntime, ndim, leafsize=LStime)
t0 = time.time()
cykdtree.PyKDTree(pts, le, re, leafsize=LStime)
t1 = time.time()
print("{} {}D points, leafsize {}: took {} s".format(
Ntime, ndim, LStime, t1 - t0))
def test_neighbors(periodic=False):
pts, le, re, ls, left_neighbors, right_neighbors = make_points_neighbors(
periodic=periodic)
tree = cykdtree.PyKDTree(pts, le, re, leafsize=ls, periodic=periodic)
for leaf in tree.leaves:
out_str = str(leaf.id)
try:
for d in range(tree.ndim):
out_str += '\nleft: {} {} {}'.format(
d, leaf.left_neighbors[d], left_neighbors[d][leaf.id])
assert (len(left_neighbors[d][leaf.id]) == len(
leaf.left_neighbors[d]))
for i in range(len(leaf.left_neighbors[d])):
assert (left_neighbors[d][leaf.id][i] ==
leaf.left_neighbors[d][i])
out_str += '\nright: {} {} {}'.format(
d, leaf.right_neighbors[d], right_neighbors[d][leaf.id])
assert (len(right_neighbors[d][leaf.id]) == len(
leaf.right_neighbors[d]))
for i in range(len(leaf.right_neighbors[d])):
assert (right_neighbors[d][leaf.id][i] ==
leaf.right_neighbors[d][i])
except:
for leaf in tree.leaves:
print(leaf.id, leaf.left_edge, leaf.right_edge)
print(out_str)
raise
def time_neighbor_search(Ntime, LStime, ndim=2):
pts, le, re, ls = make_points(Ntime, ndim, leafsize=LStime)
tree = cykdtree.PyKDTree(pts, le, re, leafsize=LStime)
t0 = time.time()
tree.get_neighbor_ids(0.5*np.ones(tree.ndim, 'double'))
t1 = time.time()
print("{} {}D points, leafsize {}: took {} s".format(Ntime, ndim, LStime, t1-t0))
def test_get_neighbor_ids(npts=100, ndim=2, periodic=False):
pts, le, re, ls = make_points(npts, ndim)
tree = cykdtree.PyKDTree(pts, le, re, leafsize=ls, periodic=periodic)
pos_list = [le, (le + re) / 2.]
if periodic:
pos_list.append(re)
for pos in pos_list:
tree.get_neighbor_ids(pos)
def test_save_load():
for periodic in (True, False):
for ndim in range(1, 5):
pts, le, re, ls = make_points(100, ndim)
tree = cykdtree.PyKDTree(pts, le, re, leafsize=ls,
periodic=periodic, data_version=ndim+12)
with tempfile.NamedTemporaryFile() as tf:
tree.save(tf.name)
restore_tree = cykdtree.PyKDTree.from_file(tf.name)
tree.assert_equal(restore_tree)
def test_PyKDTree(npts=100,
ndim=2,
periodic=False,
use_sliding_midpoint=False):
pts, le, re, ls = make_points(npts, ndim)
cykdtree.PyKDTree(pts,
le,
re,
leafsize=ls,
periodic=periodic,
use_sliding_midpoint=use_sliding_midpoint)
def test_consolidate_edges(periodic=False, ndim=2):
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
size = comm.Get_size()
pts, le, re, ls = make_points(20, ndim, leafsize=3)
Tpara = cykdtree.PyParallelKDTree(pts, le, re, leafsize=ls,
periodic=periodic)
LEpara, REpara = Tpara.consolidate_edges()
if rank == 0:
Tseri = cykdtree.PyKDTree(pts, le, re, leafsize=ls,
periodic=periodic)
LEseri, REseri = Tseri.consolidate_edges()
else:
LEseri, REseri = None, None
LEseri, REseri = comm.bcast((LEseri, REseri), root=0)
np.testing.assert_allclose(LEpara, LEseri)
np.testing.assert_allclose(REpara, REseri)
def test_consolidate(periodic=False, ndim=2):
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
size = comm.Get_size()
pts, le, re, ls = make_points(20, ndim, leafsize=3)
Tpara0 = cykdtree.PyParallelKDTree(pts, le, re, leafsize=ls,
periodic=periodic)
Tpara = Tpara0.consolidate()
if rank == 0:
if False:
from cykdtree.plot import plot2D_serial
plot2D_serial(Tpara, label_boxes=True,
plotfile='test_consolidate.png')
Tseri = cykdtree.PyKDTree(pts, le, re, leafsize=ls,
periodic=periodic)
Tpara.assert_equal(Tseri, strict_idx=False)
else:
assert(Tpara is None)
def tree(method, pts, left_edge, right_edge, periodic, *args, **kwargs):
r"""Get tree for a given domain decomposition schema.
Args:
method (str): Domain decomposition method. Supported options are:
'kdtree': KDTree based on median position along the dimension
with the greatest domain width. See
:meth:`cgal4py.domain_decomp.kdtree` for details on
accepted keyword arguments.
pts (np.ndarray of float64): (n, m) array of n coordinates in a
m-dimensional domain.
left_edge (np.ndarray of float64): (m,) domain minimum in each
dimension.
right_edge (np.ndarray of float64): (m,) domain maximum in each
dimension.
*args: Variable argument list. Passed to the selected domain
decomposition method.
**kwargs: Variable keyword argument list. Passed to the selected
domain decomposition method.
Returns:
object: Tree of type specified by `method`.
Raises:
ValueError: If `method` is not one of the supported domain
decomposition methods listed above.
"""
# Get leaves
if method.lower() == 'kdtree':
tree = kdtree.PyKDTree(pts, left_edge, right_edge, *args, **kwargs)
else:
raise ValueError("'{}' is not a supported ".format(method) +
"domain decomposition.")
# Return tree
return tree
def test_search_errors(npts=100, ndim=2):
pts, le, re, ls = make_points(npts, ndim)
tree = cykdtree.PyKDTree(pts, le, re, leafsize=ls)
assert_raises(ValueError, tree.get, re)