Tweaks for docs and an extra test

Author: benjeffery

python/tests/test_tables.py

@@ -1164,6 +1164,21 @@ def test_metadata_vector_errors(self):
         with pytest.raises(KeyError):
             _ = table.metadata_vector("x")
 
+        table.clear()
+        metadata_list = [
+            {"a": {"c": 5}, "u": [1, 2]},
+            {"a": {"b": 6}},
+        ]
+        for md in metadata_list:
+            table.add_row(
+                **{
+                    **self.input_data_for_add_row(),
+                    "metadata": md,
+                }
+            )
+        with pytest.raises(KeyError):
+            _ = table.metadata_vector(["a", "x"])
+
     def test_metadata_vector_nodefault(self):
         table = self.table_class()
         ms = tskit.MetadataSchema({"codec": "json"})

python/tskit/tables.py

@@ -49,8 +49,16 @@
 
 dataclass_options = {"frozen": True}
 
-# we'll need this because python can't tell if kwargs are passed in not
-DEFAULT = object()
+
+# Needed for cases where `None` can be an appropriate kwarg value,
+# we override the meta so that it looks good in the docs.
+class NotSetMeta(type):
+    def __repr__(cls):
+        return "Not set"
+
+
+class NOTSET(metaclass=NotSetMeta):
+    pass
 
 
 @metadata.lazy_decode
@@ -590,7 +598,7 @@ def _update_metadata_schema_cache_from_ll(self) -> None:
             self.ll_table.metadata_schema
         )
 
-    def metadata_vector(self, key, *, dtype=None, default_value=DEFAULT):
+    def metadata_vector(self, key, *, dtype=None, default_value=NOTSET):
         """
         Returns a numpy array of metadata values obtained by extracting ``key``
         from each metadata entry, and using ``default_value`` if the key is
@@ -602,10 +610,11 @@ def metadata_vector(self, key, *, dtype=None, default_value=DEFAULT):
         :param str dtype: The dtype of the result (can usually be omitted).
         :param object default_value: The value to be inserted if the metadata key
             is not present. Note that for numeric columns, a default value of None
-            will result in a non-numeric array.
+            will result in a non-numeric array. The default behaviour is to raise
+            ``KeyError`` on missing entries.
         """
 
-        if default_value == DEFAULT:
+        if default_value == NOTSET:
 
             def getter(d, k):
                 return d[k]