parcoords.py: Fix get_limits compatibility with pandas 3.0 string dtype

Pandas 3.0 now infers string columns as the dedicated string dtype instead of object dtype by default. This caused get_limits() to fail recognizing categorical columns, returning individual values instead of sets of unique values.

The fix adds a check for string dtype using pd.api.types.is_string_dtype() alongside the existing object dtype check, ensuring categorical/string columns are properly handled regardless of pandas version.

Fixes test failures in test_get_limits and test_plot when using pandas 3.0.0rc1 (--pre flag).

Author: EwoutH

ema_workbench/analysis/parcoords.py

@@ -101,7 +101,7 @@ def setup_parallel_plot(labels, minima, maxima, formatter=None, fs=14, rot=90):
 
 
 def get_limits(data):
-    """Helper function to get limits of a FataFrame that can serve as input to ParallelAxis.
+    """Helper function to get limits of a DataFrame that can serve as input to ParallelAxis.
 
     Parameters
     ----------
@@ -110,11 +110,9 @@ def get_limits(data):
     Returns
     -------
     DataFrame
-
     """
-
     def limits(x):
-        if x.dtype == "object":
+        if x.dtype == "object" or pd.api.types.is_string_dtype(x.dtype):
             return pd.Series([set(x), set(x)])
         else:
             return pd.Series([x.min(), x.max()])