Merge pull request #185 from mmaelicke/fix-nan-error

fix np.NaN usages

skgstat/MetricSpace.py

@@ -7,9 +7,9 @@
 import multiprocessing as mp
 
 
-def _sparse_dok_get(m, fill_value=np.NaN):
+def _sparse_dok_get(m, fill_value=np.nan):
     """Like m.toarray(), but setting empty values to `fill_value`, by
-    default `np.NaN`, rather than 0.0.
+    default `np.nan`, rather than 0.0.
 
     Parameters
     ----------
@@ -777,14 +777,16 @@ def dists(self):
             # from https://stackoverflow.com/questions/28677162/ignoring-duplicate-entries-in-sparse-matrix
 
             # Stable solution but a bit slow
-            # c, eq, d = zip(*set(zip(c, eq, d)))
-            # dists = sparse.csr_matrix((d, (c, eq)), shape=(len(self.coords), len(self.coords)))
+            c, eq, d = zip(*set(zip(c, eq, d)))
+            dists = sparse.csr_matrix((d, (c, eq)), shape=(len(self.coords), len(self.coords)))
 
+            # comment mmaelicke: We need to fall back to the slower solution as dok._update is not supported in scipy 1.0 anymore
+            #
             # Solution 5+ times faster than the preceding, but relies on _update() which might change in scipy (which
             # only has an implemented method for summing duplicates, and not ignoring them yet)
-            dok = sparse.dok_matrix((len(self.coords), len(self.coords)))
-            dok._update(zip(zip(c, eq), d))
-            dists = dok.tocsr()
+            # dok = sparse.dok_matrix((len(self.coords), len(self.coords)))
+            # dok._update(zip(zip(c, eq), d))
+            # dists = dok.tocsr()
 
             self._dists = dists
 

skgstat/Variogram.py

@@ -3164,9 +3164,9 @@ def __str__(self):  # pragma: no cover
         """
         par = self.describe()
 
-        _sill = np.NaN if 'error' in par else par['sill']
-        _range = np.NaN if 'error' in par else par['effective_range']
-        _nugget = np.NaN if 'error' in par else par['nugget']
+        _sill = np.nan if 'error' in par else par['sill']
+        _range = np.nan if 'error' in par else par['effective_range']
+        _nugget = np.nan if 'error' in par else par['nugget']
 
         s = "{0} Variogram\n".format(par['model'])
         s += "-" * (len(s) - 1) + "\n"

skgstat/models.py

@@ -35,7 +35,7 @@ def spherical(h, r, c0, b=0.0):
         The effective range. Note this is not the range parameter! However,
         for the spherical variogram the range and effective range are the same.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     b : float
         The nugget of the variogram. This is the value of independent
@@ -102,7 +102,7 @@ def exponential(h, r, c0, b=0.0):
         sill are exceeded. This is needed as the sill is only approached
         asymptotically by an exponential function.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     b : float
         The nugget of the variogram. This is the value of independent
@@ -164,7 +164,7 @@ def gaussian(h, r, c0, b=0.0):
         sill are exceeded. This is needed as the sill is only approached
         asymptotically by an exponential function.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     b : float
         The nugget of the variogram. This is the value of independent
@@ -226,7 +226,7 @@ def cubic(h, r, c0, b=0.0):
         The effective range. Note this is not the range parameter! However,
         for the cubic variogram the range and effective range are the same.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     b : float
         The nugget of the variogram. This is the value of independent
@@ -297,7 +297,7 @@ def stable(h, r, c0, s, b=0.0):
         only approached asymptotically by the e-function part of the stable
         model.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     s : float
         Shape parameter. For s <= 2 the model will be shaped more like a
@@ -369,7 +369,7 @@ def matern(h, r, c0, s, b=0.0):
         where 95% of the sill are exceeded. This is needed as the sill is
         only approached asymptotically by Matérn model.
     c0 : float
-        The sill of the variogram, where it will flatten out. The function
+        The partial sill of the variogram, where it will flatten out. The function
         will not return a value higher than C0 + b.
     s : float
         Smoothness parameter. The smoothness parameter can shape a smooth or

skgstat/tests/test_variogram.py

@@ -896,7 +896,7 @@ def test_harmonize_model(self):
 
         assert_array_almost_equal(
             V.transform(x),
-            [np.NaN, 0.57, 1.01, 1.12, 1.15, 1.15, 1.15, 1.15, 1.21, 1.65],
+            [np.nan, 0.57, 1.01, 1.12, 1.15, 1.15, 1.15, 1.15, 1.21, 1.65],
             decimal=2
         )