Add an explicit test for Dataset.graph without identifier

This is to test the somewhat arbitrary behaviour of binding the `genid`
namespace when `Dataset.graph` is used without an identifier.

test/test_dataset/test_dataset.py

@@ -1,13 +1,14 @@
 # -*- coding: utf-8 -*-
 import os
+import re
 import shutil
 import tempfile
 from test.data import context1, likes, pizza, tarek
 
 import pytest
 
 from rdflib import URIRef, plugin
-from rdflib.graph import DATASET_DEFAULT_GRAPH_ID, Dataset
+from rdflib.graph import DATASET_DEFAULT_GRAPH_ID, Dataset, Graph, Namespace
 
 # Will also run SPARQLUpdateStore tests against local SPARQL1.1 endpoint if
 # available. This assumes SPARQL1.1 query/update endpoints running locally at
@@ -231,3 +232,26 @@ def test_iter(get_dataset):
         i_new += 1
 
     assert i_new == i_trad  # both should be 3
+
+
+EGSCHEMA = Namespace("example:")
+
+
+def test_subgraph_without_identifier() -> None:
+    """
+    Graphs with no identifies assigned are identified by Skolem IRIs with a
+    prefix that is bound to `genid`.
+
+    TODO: This is somewhat questionable and arbitrary behaviour and should be
+    reviewed at some point.
+    """
+
+    dataset = Dataset()
+
+    subgraph: Graph = dataset.graph()
+    subgraph.add((EGSCHEMA["subject"], EGSCHEMA["predicate"], EGSCHEMA["object"]))
+
+    data = dataset.serialize(format="trig")
+
+    re.match(r"\s+genid:\s+", data, re.MULTILINE | re.DOTALL)
+    re.match(r"genid:\S", data, re.MULTILINE | re.DOTALL)