Merge branch 'charnley/consistency' into charnley/cites

.coveragerc

@@ -0,0 +1,5 @@
+[report]
+exclude_also =
+    def __repr__
+    raise ValueError
+    raise NotImplementedError

Makefile

@@ -34,7 +34,7 @@ types:
 	${python} -m monkeytype list-modules | grep ${pkg} | parallel -j${j} "${python} -m monkeytype apply {} > /dev/null && echo {}"
 
 cov:
-	${python} -m pytest -vrs --cov=${pkg} --cov-report html tests
+	${python} -m pytest --cov=${pkg} --cov-config .coveragerc --cov-report html tests
 
 compile:
 	${python} _compile.py

environment_dev.yaml

@@ -14,6 +14,7 @@ dependencies:
     - pre-commit
     - pyarrow
     - pytest
+    - pytest-cov
     - scikit-learn
     - scipy
     # build

src/qmllib/kernels/fdistance.f90

@@ -1,4 +1,3 @@
-
 subroutine fmanhattan_distance(A, B, D)
 
    implicit none

src/qmllib/kernels/fgradient_kernels.f90

@@ -1,25 +1,3 @@
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
 subroutine fglobal_kernel(x1, x2, q1, q2, n1, n2, nm1, nm2, sigma, kernel)
 
     implicit none

src/qmllib/kernels/fkernels.f90

@@ -1,4 +1,3 @@
-
 subroutine fget_local_kernels_gaussian(q1, q2, n1, n2, sigmas, &
         & nm1, nm2, nsigmas, kernels)
 

src/qmllib/kernels/fkpca.f90

@@ -1,4 +1,3 @@
-
 subroutine fkpca(k, n, centering, kpca)
 
    implicit none

src/qmllib/kernels/fkwasserstein.f90

@@ -1,4 +1,3 @@
-
 module searchtools
 
    implicit none

src/qmllib/kernels/gradient_kernels.py

@@ -59,12 +59,10 @@ def get_global_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("Error: List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -114,12 +112,10 @@ def get_local_kernels(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("Error: List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("Error: List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -176,12 +172,10 @@ def get_local_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -228,9 +222,8 @@ def get_local_symmetric_kernels(X1: ndarray, Q1: List[List[int]], SIGMAS: List[f
 
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("Error: List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     for i, q in enumerate(Q1):
@@ -275,9 +268,8 @@ def get_local_symmetric_kernel(
 
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("Error: List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     for i, q in enumerate(Q1):
@@ -329,12 +321,10 @@ def get_atomic_local_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -394,12 +384,10 @@ def get_atomic_local_gradient_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -475,12 +463,10 @@ def get_local_gradient_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -552,12 +538,10 @@ def get_gdml_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -627,9 +611,8 @@ def get_symmetric_gdml_kernel(
 
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
 
@@ -692,12 +675,10 @@ def get_gp_kernel(
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
     N2 = np.array([len(Q) for Q in Q2], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
-    assert (
-        N2.shape[0] == X2.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
+    if not (N2.shape[0] == X2.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
     Q2_input = np.zeros((max(N2), X2.shape[0]), dtype=np.int32)
@@ -765,9 +746,8 @@ def get_symmetric_gp_kernel(
 
     N1 = np.array([len(Q) for Q in Q1], dtype=np.int32)
 
-    assert (
-        N1.shape[0] == X1.shape[0]
-    ), "Error: List of charges does not match shape of representations"
+    if not (N1.shape[0] == X1.shape[0]):
+        raise ValueError("List of charges does not match shape of representations")
 
     Q1_input = np.zeros((max(N1), X1.shape[0]), dtype=np.int32)
 

src/qmllib/kernels/kernels.py

@@ -265,25 +265,28 @@ def matern_kernel(
     """
 
     if metric == "l1":
+
         if order == 0:
             gammas = []
+
         elif order == 1:
             gammas = [1]
             sigma /= np.sqrt(3)
+
         elif order == 2:
             gammas = [1, 1 / 3.0]
             sigma /= np.sqrt(5)
+
         else:
-            print("Order:%d not implemented in Matern Kernel" % order)
-            raise SystemExit
+            raise ValueError(f"Order '{order}' not implemented in Matern Kernel")
 
         return sargan_kernel(A, B, sigma, gammas)
 
     elif metric == "l2":
         pass
+
     else:
-        print("Error: Unknown distance metric %s in Matern kernel" % str(metric))
-        raise SystemExit
+        raise ValueError(f"Unknown distance metric {metric} in Matern kernel")
 
     na = A.shape[0]
     nb = B.shape[0]
@@ -326,10 +329,13 @@ def get_local_kernels_gaussian(
     :rtype: numpy array
     """
 
-    assert np.sum(na) == A.shape[0], "Error in A input"
-    assert np.sum(nb) == B.shape[0], "Error in B input"
+    if np.sum(na) != A.shape[0]:
+        raise ValueError("Error in A input")
+    if np.sum(nb) != B.shape[0]:
+        raise ValueError("Error in B input")
 
-    assert A.shape[1] == B.shape[1], "Error in representation sizes"
+    if A.shape[1] != B.shape[1]:
+        raise ValueError("Error in representation sizes")
 
     nma = len(na)
     nmb = len(nb)
@@ -370,10 +376,13 @@ def get_local_kernels_laplacian(
     :rtype: numpy array
     """
 
-    assert np.sum(na) == A.shape[0], "Error in A input"
-    assert np.sum(nb) == B.shape[0], "Error in B input"
+    if np.sum(na) != A.shape[0]:
+        raise ValueError("Error in A input")
+    if np.sum(nb) != B.shape[0]:
+        raise ValueError("Error in B input")
 
-    assert A.shape[1] == B.shape[1], "Error in representation sizes"
+    if A.shape[1] != B.shape[1]:
+        raise ValueError("Error in representation sizes")
 
     nma = len(na)
     nmb = len(nb)
@@ -403,9 +412,12 @@ def kpca(K: ndarray, n: int = 2, centering: bool = True) -> ndarray:
     :rtype: numpy array
     """
 
-    assert K.shape[0] == K.shape[1], "ERROR: Square matrix required for Kernel PCA."
-    assert np.allclose(K, K.T, atol=1e-8), "ERROR: Symmetric matrix required for Kernel PCA."
-    assert n <= K.shape[0], "ERROR: Requested more principal components than matrix size."
+    if K.shape[0] != K.shape[1]:
+        raise ValueError("Square matrix required for Kernel PCA.")
+    if not np.allclose(K, K.T, atol=1e-8):
+        raise ValueError("Symmetric matrix required for Kernel PCA.")
+    if not n <= K.shape[0]:
+        raise ValueError("Requested more principal components than matrix size.")
 
     size = K.shape[0]
     pca = fkpca(K, size, centering)