BUG: fix infer_dtype for StringDtype

doc/source/whatsnew/v1.0.2.rst

@@ -32,6 +32,11 @@ Bug fixes
 
 - Using ``pd.NA`` with :meth:`DataFrame.to_json` now correctly outputs a null value instead of an empty object (:issue:`31615`)
 
+
+**Experimental dtypes**
+
+- Fix bug in :meth:`DataFrame.convert_dtypes` for columns that were already using the ``"string"`` dtype (:issue:`31731`).
+
 .. ---------------------------------------------------------------------------
 
 .. _whatsnew_102.contributors:

pandas/_libs/lib.pyx

@@ -1005,7 +1005,7 @@ _TYPE_MAP = {
     'complex64': 'complex',
     'complex128': 'complex',
     'c': 'complex',
-    'string': 'bytes',
+    'string': 'string',
     'S': 'bytes',
     'U': 'string',
     'bool': 'boolean',

pandas/tests/dtypes/test_inference.py

@@ -1200,6 +1200,20 @@ def test_interval(self):
         inferred = lib.infer_dtype(pd.Series(idx), skipna=False)
         assert inferred == "interval"
 
+    def test_string_dtype(self):
+        # StringArray
+        arr = pd.array(["a", "b", pd.NA], dtype="string")
+        for val in [list(arr), arr, pd.Series(arr)]:
+            inferred = lib.infer_dtype(val)
+            assert inferred == "string"
+
+    def test_boolean_dtype(self):
+        # BooleanArray
+        arr = pd.array([True, False, pd.NA], dtype="boolean")
+        for val in [list(arr), arr, pd.Series(arr)]:
+            inferred = lib.infer_dtype(val)
+            assert inferred == "boolean"
+
 
 class TestNumberScalar:
     def test_is_number(self):

pandas/tests/series/methods/test_convert_dtypes.py

@@ -246,3 +246,10 @@ def test_convert_dtypes(self, data, maindtype, params, answerdict):
 
         # Make sure original not changed
         tm.assert_series_equal(series, copy)
+
+    def test_convert_string_dtype(self):
+        # https://github.com/pandas-dev/pandas/issues/31731 -> converting columns
+        # that are already string dtype
+        df = pd.DataFrame({"A": ["a", "b", "c"], "B": ["ä", "ö", "ü"]}, dtype="string")
+        result = df.convert_dtypes()
+        tm.assert_frame_equal(df, result)