guess %Y-%m format

Author: MarcoGorelli

pandas/_libs/tslibs/parsing.pyx

@@ -1011,10 +1011,11 @@ def guess_datetime_format(dt_str: str, bint dayfirst=False) -> str | None:
                 break
 
     # Only consider it a valid guess if we have a year, month and day,
-    # unless it's %Y which is both common and unambiguous.
+    # unless it's %Y or %Y-%m which conform with ISO8601. Note that we don't
+    # make an exception for %Y%m because it's explicitly not considered ISO8601.
     if (
         len({'year', 'month', 'day'} & found_attrs) != 3
-        and format_guess != ['%Y']
+        and format_guess not in (['%Y'], ['%Y', None, '%m'])
     ):
         return None
 

pandas/tests/tslibs/test_parsing.py

@@ -148,6 +148,7 @@ def test_parsers_month_freq(date_str, expected):
         ("20111230", "%Y%m%d"),
         ("2011-12-30", "%Y-%m-%d"),
         ("2011", "%Y"),
+        ("2011-01", "%Y-%m"),
         ("30-12-2011", "%d-%m-%Y"),
         ("2011-12-30 00:00:00", "%Y-%m-%d %H:%M:%S"),
         ("2011-12-30T00:00:00", "%Y-%m-%dT%H:%M:%S"),
@@ -215,6 +216,7 @@ def test_guess_datetime_format_with_locale_specific_formats(string, fmt):
         "this_is_not_a_datetime",
         "51a",
         "13/2019",
+        "202001",  # YYYYMM isn't ISO8601
     ],
 )
 def test_guess_datetime_format_invalid_inputs(invalid_dt):