BUG: to_period() freq was not infered

pandas/core/arrays/datetimes.py

@@ -18,6 +18,7 @@
     timezones,
     tzconversion,
 )
+import pandas._libs.tslibs.frequencies as libfrequencies
 from pandas.errors import PerformanceWarning
 
 from pandas.core.dtypes.common import (
@@ -1097,7 +1098,13 @@ def to_period(self, freq=None):
                     "You must pass a freq argument as current index has none."
                 )
 
-            freq = get_period_alias(freq)
+            res = get_period_alias(freq)
+
+            if res is None:
+                base, stride = libfrequencies._base_and_stride(freq)
+                res = f"{stride}{base}"
+
+            freq = res
 
         return PeriodArray._from_datetime64(self._data, freq, tz=self.tz)
 

pandas/tests/indexes/datetimes/test_to_period.py

@@ -1,3 +1,5 @@
+import warnings
+
 import dateutil.tz
 from dateutil.tz import tzlocal
 import pytest
@@ -75,6 +77,22 @@ def test_to_period_monthish(self):
         with pytest.raises(ValueError, match=INVALID_FREQ_ERR_MSG):
             date_range("01-Jan-2012", periods=8, freq="EOM")
 
+    def test_to_period_infer(self):
+        # https://github.com/pandas-dev/pandas/issues/33358
+        rng = date_range(
+            start="2019-12-22 06:40:00+00:00",
+            end="2019-12-22 08:45:00+00:00",
+            freq="5min",
+        )
+
+        # Using simple filter because we are not checking for the warning here
+        warnings.simplefilter("ignore", UserWarning)
+
+        pi1 = rng.to_period("5min")
+        pi2 = rng.to_period()
+
+        tm.assert_index_equal(pi1, pi2)
+
     def test_period_dt64_round_trip(self):
         dti = date_range("1/1/2000", "1/7/2002", freq="B")
         pi = dti.to_period()