Fixed pystan testing issue.

pints/interfaces/stan/_stan.py

@@ -8,8 +8,9 @@
 import re
 import warnings
 
-import httpstan
 import numpy as np
+
+import httpstan
 import stan
 
 import pints
@@ -20,9 +21,9 @@ class StanLogPDF(pints.LogPDF):
     A :class:`pints.LogPDF` based on Stan code and data, which can be used in
     sampling and optimisation.
 
-    This class interfaces with Stan ([1]_) using PyStan ([2]_) and httpspan
-    ([3]_), which compiles the Stan model code. This can take some time
-    (typically minutes).
+    This class interfaces with Stan ([1]_) using ``stan`` ([2]_) and
+    ``httpspan`` ([3]_), which compiles the Stan model code. This can take some
+    time (typically minutes).
 
     Note that the interface assumes that the parameters are on the
     unconstrained scale (according to Stan's "constraint transforms" [1]_).
@@ -32,7 +33,7 @@ class StanLogPDF(pints.LogPDF):
     back to the constrained space (in the example, using an ``exp`` transform)
     to obtain appropriate samples.
 
-    Parameters are ordered as in the stan model. Vector and matrix parameters
+    Parameters are ordered as in the Stan model. Vector and matrix parameters
     are "flattened" into a sequence by Stan, use :meth:`StanLogPDF.names()` to
     see the result.
 

pints/tests/test_interfaces_stan.py

@@ -15,16 +15,18 @@
 from shared import SubCapture
 
 
+debug = False
+
+no_stan_msg = None
 try:
     from pints.interfaces.stan import StanLogPDF
     have_stan = True
-except ImportError:
+except ImportError as e:
     have_stan = False
-
-debug = False
+    no_stan_msg = f'PyStan not found: {e}'
 
 
-@unittest.skipIf(not have_stan, 'PyStan not found')
+@unittest.skipIf(not have_stan, no_stan_msg)
 class TestStanLogPDF(unittest.TestCase):
     """ Tests StanLogPDF. """
 

setup.py

@@ -82,6 +82,7 @@ def load_version():
         ],
         'stan': [
             'pystan>=3',
+            'setuptools<82',
         ]
     },
     python_requires='>=3.8',