REGR: Fix construction of PeriodIndex from strings

doc/source/whatsnew/v1.1.0.rst

@@ -305,6 +305,7 @@ Datetimelike
 - :class:`Timestamp` raising confusing error message when year, month or day is missing (:issue:`31200`)
 - Bug in :class:`DatetimeIndex` constructor incorrectly accepting ``bool``-dtyped inputs (:issue:`32668`)
 - Bug in :meth:`DatetimeIndex.searchsorted` not accepting a ``list`` or :class:`Series` as its argument (:issue:`32762`)
+- Bug where :meth:`PeriodIndex` raised when passed a :class:`Series` of strings (:issue:`26109`)
 
 Timedelta
 ^^^^^^^^^

pandas/core/arrays/period.py

@@ -27,6 +27,7 @@
     ensure_object,
     is_datetime64_dtype,
     is_float_dtype,
+    is_list_like,
     is_period_dtype,
     pandas_dtype,
 )
@@ -831,11 +832,11 @@ def period_array(
     """
     if is_datetime64_dtype(data):
         return PeriodArray._from_datetime64(data, freq)
-    if isinstance(data, (ABCPeriodIndex, ABCSeries, PeriodArray)):
+    if is_period_dtype(data):
         return PeriodArray(data, freq)
 
     # other iterable of some kind
-    if not isinstance(data, (np.ndarray, list, tuple)):
+    if not is_list_like(data):
         data = list(data)
 
     data = np.asarray(data)

pandas/tests/arrays/test_datetimelike.py

@@ -868,3 +868,10 @@ def test_searchsorted_datetimelike_with_listlike_invalid_dtype(values, arg):
     msg = "[Unexpected type|Cannot compare]"
     with pytest.raises(TypeError, match=msg):
         values.searchsorted(arg)
+
+
+@pytest.mark.parametrize("klass", [list, tuple, np.array, pd.Series])
+def test_period_index_construction_from_strings(klass):
+    result = PeriodIndex(klass(["2020Q1", "2020Q2", "2020Q3", "2020Q4"]), freq="Q")
+    expected = pd.period_range(start="2020", periods=4, freq="Q")
+    tm.assert_index_equal(result, expected)