BUG: closes #44312: fixes unwanted TypeError when a missing metadata field is missing (#44325)

Author: qdbp

doc/source/whatsnew/v1.4.0.rst

@@ -632,6 +632,7 @@ I/O
 - Bug in :func:`read_csv`, changed exception class when expecting a file path name or file-like object from ``OSError`` to ``TypeError`` (:issue:`43366`)
 - Bug in :func:`read_json` not handling non-numpy dtypes correctly (especially ``category``) (:issue:`21892`, :issue:`33205`)
 - Bug in :func:`json_normalize` where multi-character ``sep`` parameter is incorrectly prefixed to every key (:issue:`43831`)
+- Bug in :func:`json_normalize` where reading data with missing multi-level metadata would not respect errors="ignore" (:issue:`44312`)
 - Bug in :func:`read_csv` with :code:`float_precision="round_trip"` which did not skip initial/trailing whitespace (:issue:`43713`)
 - Bug in dumping/loading a :class:`DataFrame` with ``yaml.dump(frame)`` (:issue:`42748`)
 - Bug in :func:`read_csv` raising ``ValueError`` when ``parse_dates`` was used with ``MultiIndex`` columns (:issue:`8991`)

pandas/io/json/_normalize.py

@@ -389,6 +389,8 @@ def _pull_field(
         try:
             if isinstance(spec, list):
                 for field in spec:
+                    if result is None:
+                        raise KeyError(field)
                     result = result[field]
             else:
                 result = result[spec]

pandas/tests/io/json/test_normalize.py

@@ -634,6 +634,33 @@ def test_missing_meta(self, missing_metadata):
         expected = DataFrame(ex_data, columns=columns)
         tm.assert_frame_equal(result, expected)
 
+    def test_missing_nested_meta(self):
+        # GH44312
+        # If errors="ignore" and nested metadata is null, we should return nan
+        data = {"meta": "foo", "nested_meta": None, "value": [{"rec": 1}, {"rec": 2}]}
+        result = json_normalize(
+            data,
+            record_path="value",
+            meta=["meta", ["nested_meta", "leaf"]],
+            errors="ignore",
+        )
+        ex_data = [[1, "foo", np.nan], [2, "foo", np.nan]]
+        columns = ["rec", "meta", "nested_meta.leaf"]
+        expected = DataFrame(ex_data, columns=columns).astype(
+            {"nested_meta.leaf": object}
+        )
+        tm.assert_frame_equal(result, expected)
+
+        # If errors="raise" and nested metadata is null, we should raise with the
+        # key of the first missing level
+        with pytest.raises(KeyError, match="'leaf' not found"):
+            json_normalize(
+                data,
+                record_path="value",
+                meta=["meta", ["nested_meta", "leaf"]],
+                errors="raise",
+            )
+
     def test_missing_meta_multilevel_record_path_errors_raise(self, missing_metadata):
         # GH41876
         # Ensure errors='raise' works as intended even when a record_path of length