Fix issue with mypy and use ubuntu-20.04 for testing

Python 3.7 is not supported anymore in the latest
ubuntu used in the GitHub action; use ubuntu-20.04
instead.

The package is now tested on Python v3.12 and v3.13.

.github/workflows/main.yml

@@ -42,7 +42,7 @@ jobs:
         run: tox -e typecheck
   test:
     name: Test
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04
     strategy:
       matrix:
         python:
@@ -54,6 +54,10 @@ jobs:
             toxenv: "py39"
           - version: "3.11"
             toxenv: "py311"
+          - version: "3.12"
+            toxenv: "py312"
+          - version: "3.13"
+            toxenv: "py313"
     steps:
       - uses: actions/checkout@v3
       - uses: actions/setup-python@v4.0.0

src/uqtestfuns/core/prob_input/univariate_distributions/triangular.py

@@ -6,7 +6,7 @@
 These are the lower bound, upper bound, and the mid-point (i.e., modal value),
 respectively.
 
-The density value at the modal value is computed such that the are under
+The density value at the modal value is computed such that the area under
 the distribution is 1.0.
 """
 
@@ -238,6 +238,6 @@ def icdf(
     yy[idx_nan] = np.nan
 
     # Check if values are within the set bounds
-    yy = postprocess_icdf(yy, lower_bound, upper_bound)
+    yy_post = postprocess_icdf(yy, lower_bound, upper_bound)
 
-    return yy
+    return yy_post

src/uqtestfuns/core/prob_input/univariate_distributions/uniform.py

@@ -201,6 +201,6 @@ def icdf(
     yy = lower_bound + np.diff(parameters) * xx
 
     # Check if values are within the set bounds
-    yy = postprocess_icdf(yy, lower_bound, upper_bound)
+    yy_post = postprocess_icdf(yy, lower_bound, upper_bound)
 
-    return yy
+    return yy_post

tests/core/prob_input/test_univariate_input.py

@@ -26,13 +26,15 @@ def univariate_input(
     # to avoid awkward yet unrealistic values)
     name = create_random_alphanumeric(8)
     distribution = request.param
-    if request.param == "uniform":
+    if distribution == "uniform":
         parameters = np.sort(np.round(np.random.rand(2), decimals=5))
-    elif request.param == "beta":
+    elif distribution == "beta":
         parameters = np.sort(np.round(np.random.rand(4), decimals=5))
     elif distribution == "exponential":
         # Single parameter, must be strictly positive
-        parameters = 1 + np.round(np.random.rand(1), decimals=5)
+        parameters = (1 + np.round(np.random.rand(1), decimals=5)).astype(
+            np.float64
+        )
     elif distribution == "triangular":
         parameters = np.sort(1 + 2 * np.round(np.random.rand(2), decimals=5))
         # Append the mid point
@@ -49,9 +51,13 @@ def univariate_input(
         parameters = np.insert(parameters, 1, np.random.rand(1))
     elif distribution == "lognormal":
         # Limit the size of the parameters
-        parameters = 1 + np.round(np.random.rand(2), decimals=5)
+        parameters = (1 + np.round(np.random.rand(2), decimals=5)).astype(
+            np.float64
+        )
     else:
-        parameters = 5 * np.round(np.random.rand(2), decimals=5)
+        parameters = (5 * np.round(np.random.rand(2), decimals=5)).astype(
+            np.float64
+        )
         parameters[1] += 1.0
 
     specs = {