Json normalize nan support

doc/source/whatsnew/v0.25.0.rst

@@ -214,7 +214,7 @@ I/O
 - Bug in :func:`read_json` for ``orient='table'`` when it tries to infer dtypes by default, which is not applicable as dtypes are already defined in the JSON schema (:issue:`21345`)
 - Bug in :func:`read_json` for ``orient='table'`` and float index, as it infers index dtype by default, which is not applicable because index dtype is already defined in the JSON schema (:issue:`25433`)
 - Bug in :func:`read_json` for ``orient='table'`` and string of float column names, as it makes a column name type conversion to Timestamp, which is not applicable because column names are already defined in the JSON schema (:issue:`25435`)
--
+- Bug in :func:`json_normalize` for ``errors='ignore'`` and nullable metadata fields, the null values in dataframe were literal nan string and not numpy.nan (:issue:`25468`)
 -
 -
 

pandas/io/json/normalize.py

@@ -281,6 +281,6 @@ def _recursive_extract(data, path, seen_meta, level=0):
             raise ValueError('Conflicting metadata name {name}, '
                              'need distinguishing prefix '.format(name=k))
 
-        result[k] = np.array(v).repeat(lengths)
+        result[k] = np.array(v, dtype=object).repeat(lengths)
 
     return result

pandas/tests/io/json/test_normalize.py

@@ -66,6 +66,25 @@ def author_missing_data():
          }]
 
 
+@pytest.fixture
+def address_missing_data():
+    return [
+        {'name': 'Alice',
+         'addresses': [{'number': 9562,
+                        'street': 'Morris St.',
+                        'city': 'Massillon',
+                        'state': 'OH',
+                        'zip': 44646}]
+         },
+        {'addresses': [{'number': 8449,
+                        'street': 'Spring St.',
+                        'city': 'Elizabethton',
+                        'state': 'TN',
+                        'zip': 37643}]
+         }
+    ]
+
+
 class TestJSONNormalize(object):
 
     def test_simple_records(self):
@@ -378,6 +397,31 @@ def test_json_normalize_errors(self):
                       ['general', 'trade_version']],
                 errors='raise')
 
+    def test_missing_meta(self, address_missing_data):
+        # GH25468: If metadata is nullable with errors set to ignore, the null
+        # values should be numpy.nan values
+        result = json_normalize(
+            data=address_missing_data,
+            record_path='addresses',
+            meta='name',
+            errors='ignore')
+        ex_data = [
+            {'city': 'Massillon',
+             'number': 9562,
+             'state': 'OH',
+             'street': 'Morris St.',
+             'zip': 44646,
+             'name': 'Alice'},
+            {'city': 'Elizabethton',
+             'number': 8449,
+             'state': 'TN',
+             'street': 'Spring St.',
+             'zip': 37643,
+             'name': np.nan}
+        ]
+        expected = DataFrame(ex_data)
+        tm.assert_frame_equal(result, expected, check_like=True)
+
     def test_donot_drop_nonevalues(self):
         # GH21356
         data = [