pandas version-dependent map or applymap

It seems the transition from `applymap` to `map` is <1 year old,
starting pandas version 2.1:
https://pandas.pydata.org/pandas-docs/version/2.1/reference/api/pandas.DataFrame.map.html

This is too new to fully rely on.
Especially when it breaks for python 3.7 and 3.8
that are not supported by newer pandas versions,
which I'm not yet comfortable with (although their end-of-life status,
because I'm not sure how causallib supports 3.12 at the moment).

So putting a pandas version-dependent choice makes sense at the moment
and can later be discarded.

Signed-off-by: Ehud-Karavani <[email protected]>

Author: ehudkr

causallib/tests/test_matching.py

@@ -542,11 +542,20 @@ def test_classify_task(self):
         ypred = self.matching.estimate_individual_outcome(
             X, a, predict_proba=False)
         # if not all integers will raise errors
-        ypred.map(str).map(int)
-        with self.assertRaises(ValueError):
-            self.matching.estimate_individual_outcome(
-                X, a, predict_proba=True
-            ).map(str).map(int)
+        if pd.__version__ > "2.1":
+            # TODO: at time of writing pandas 2.1.0 is <1 year old.
+            #       Once matured, you may erase the deprecated `applymap` and just use `map`.
+            ypred.map(str).map(int)
+            with self.assertRaises(ValueError):
+                self.matching.estimate_individual_outcome(
+                    X, a, predict_proba=True
+                ).map(str).map(int)
+        else:
+            ypred.applymap(str).applymap(int)
+            with self.assertRaises(ValueError):
+                self.matching.estimate_individual_outcome(
+                    X, a, predict_proba=True
+                ).applymap(str).applymap(int)
 
     def test_classify_task_with_wrong_inputs_warning_check(self):
         X, a, y = self.data_3feature_linear_effect