MAINT: avoid warning for Pandas >= 2.0

gdementen · gdementen · commit 92dfa8612d4e · 2023-09-13T15:40:59.000+02:00
some stricter (it only ever infers once) version of infer_datetime_format is now the default
I choose not to bump pandas requirements to 2.0 because as far as I can tell, this only changes performance for metadata (little used feature) on Pandas &lt; 2
diff --git a/larray/core/metadata.py b/larray/core/metadata.py
@@ -66,9 +66,11 @@ def from_array(cls, array) -> 'Metadata':
         from pandas import to_numeric, to_datetime
 
         def _convert_value(value):
+            # errors='ignore' => the value is unmodified if it does not parse
             value = to_numeric([value], errors='ignore')[0]
             if isinstance(value, str):
-                value = to_datetime(value, errors='ignore', infer_datetime_format=True)
+                # same here
+                value = to_datetime(value, errors='ignore')
             return value
 
         return Metadata({key: _convert_value(value) for key, value in zip(array.axes.labels[0], array.data)})
