def sans_benchmark(request):
r"""Sassena output containing 1000 I(Q) profiles for the hiAPP centroids.
Yields
------
dict
'profiles' : HDF5 handle to the file containing the I(Q) profiles
'property_list' : list of SansProperty instances, one for each leaf
'tree_with_no_property' : cnextend.Tree with random distances among
leafs and without included properties.
"""
# setup or initialization
handle = h5py.File(os.path.join(data_dir, 'sans', 'profiles.h5'), 'r')
profiles = handle['fqt']
n_leafs = len(profiles)
# Create a node tree.
# m is a 1D compressed matrix of distances between leafs
m = np.random.random(int(n_leafs * (n_leafs - 1) / 2))
z = linkage(m)
tree = cnextend.Tree(z)
# values is a list of SansProperty instances, one for each tree leaf
values = list()
for i in range(tree.nleafs):
sans_property = idprop.SansProperty()
sans_property.from_sassena(handle, index=i)
values.append(sans_property)
def teardown():
handle.close()
request.addfinalizer(teardown)
return dict(profiles=handle, property_list=values,
tree_with_no_property=tree)
def benchmark():
z = np.loadtxt(os.path.join(data_dir, 'linkage_matrix'))
return {'z': z,
'tree': cnextend.Tree(z),
'nnodes': 44757,
'nleafs': 22379,
'simple_property': [SimpleProperty(i) for i in range(22379)],
}
def test_from_linkage_matrix(self, benchmark):
t = cnx.Tree()
t.from_linkage_matrix(benchmark['z'], node_class=hierarchy.ClusterNode)
r = t.root
assert hasattr(r, 'parent') is False
t.from_linkage_matrix(benchmark['z'], node_class=cnx.ClusterNodeX)
r = t.root
assert r.parent is None
assert len(t) == benchmark['nnodes']
def small_tree():
n_leafs = 9
a = np.arange(n_leafs)
dist_mat = squareform(np.square(a - a[:, np.newaxis]))
z = linkage(dist_mat, method='complete')
return {'dist_mat': dist_mat,
'z': z,
'tree': cnextend.Tree(z),
'simple_property': [SimpleProperty(i) for i in range(n_leafs)],
}
def benchmark():
z = np.loadtxt(os.path.join(data_dir, 'linkage_matrix'))
t = cnextend.Tree(z)
n_leafs = 22379
# Instantiate scalar properties for the leaf nodes, then propagate
# up the tree
sc = np.random.normal(loc=100.0, size=n_leafs)
sc_p = [idprop.ScalarProperty(name='sc', y=s) for s in sc]
idprop.propagator_size_weighted_sum(sc_p, t)
return {
'z': z,
'tree': t,
'nnodes': 44757,
'nleafs': n_leafs,
'simple_property': [SimpleProperty(i) for i in range(22379)],
}