fix: compat helper for DataFrame map vs applymap across pandas versions

pandas 2.1+ removed applymap; Python 3.8 CI cannot install pandas>=2.1.
Use applymap when present, else map.

Author: coreyneskey

pyfair/report/base_report.py

@@ -22,6 +22,14 @@
 from ..utility.beta_pert import FairBetaPert
 
 
+def _dataframe_elementwise(df, func):
+    """Element-wise transform; pandas 2.1+ uses map, older uses applymap."""
+    applymap = getattr(df, "applymap", None)
+    if applymap is not None:
+        return applymap(func)
+    return df.map(func)
+
+
 class FairBaseReport(object):
     """A base class for creating FairModel and FairMetaModel reports
 
@@ -262,8 +270,8 @@ def _get_overview_table(self, model_or_models):
         risk_results = risk_results.agg(["mean", "std", "min", "max"])
         risk_results.index = ["Mean", "Stdev", "Minimum", "Maximum"]
         # Format risk results into dataframe
-        overview_df = risk_results.map(
-            lambda x: self._format_strings["Risk"].format(x)
+        overview_df = _dataframe_elementwise(
+            risk_results, lambda x: self._format_strings["Risk"].format(x)
         )
         overview_df.loc["Simulations"] = [
             "{0:,.0f}".format(len(model.export_results()))
@@ -321,7 +329,9 @@ def _get_model_parameter_table(self, model):
             # On a column basis
             axis=1,
         )
-        param_df = param_df.map(lambda x: "" if "nan" in x else x)
+        param_df = _dataframe_elementwise(
+            param_df, lambda x: "" if "nan" in x else x
+        )
         # Do not truncate our base64 images.
         pd.set_option("display.max_colwidth", None)
         # Create our distribution icons as strings in table

requirements.txt

@@ -1,4 +1,4 @@
-pandas>=2.1.0
+pandas>=0.24.1
 numpy>=1.16.1
 scipy>=1.2.1
 matplotlib>=3.0.2