From 94cdca85780a5b3616b91683773f75b85e4fe2ea Mon Sep 17 00:00:00 2001
From: Edmond Chuc <edmond@kurrawong.ai>
Date: Wed, 3 Sep 2025 16:38:33 +1000
Subject: [PATCH 1/2] feat: canonicalization with longturtle serializer now
 optional

Fixes https://github.com/RDFLib/rdflib/issues/3196
---
 rdflib/plugins/serializers/longturtle.py      | 45 ++++++++++++++-----
 .../test_serializer_longturtle.py             |  2 +-
 .../test_serializer_longturtle_sort.py        |  2 +-
 3 files changed, 35 insertions(+), 14 deletions(-)

diff --git a/rdflib/plugins/serializers/longturtle.py b/rdflib/plugins/serializers/longturtle.py
index 95b94efd7..1315181b1 100644
--- a/rdflib/plugins/serializers/longturtle.py
+++ b/rdflib/plugins/serializers/longturtle.py
@@ -39,23 +39,20 @@
 
 
 class LongTurtleSerializer(RecursiveSerializer):
+    """LongTurtle, a Turtle serialization format.
+
+    When the optional parameter `canon` is set to True, the graph is canonicalized
+    before serialization. This normalizes blank node identifiers and allows for
+    deterministic serialization of the graph. Useful when consistent outputs are required.
+    """
+
     short_name = "longturtle"
     indentString = "    "
 
     def __init__(self, store):
         self._ns_rewrite = {}
-        namespace_manager = store.namespace_manager
-        store = to_canonical_graph(store)
-        content = store.serialize(format="application/n-triples")
-        lines = content.split("\n")
-        lines.sort()
-        graph = Graph()
-        graph.parse(
-            data="\n".join(lines), format="application/n-triples", skolemize=True
-        )
-        graph = graph.de_skolemize()
-        graph.namespace_manager = namespace_manager
-        super(LongTurtleSerializer, self).__init__(graph)
+        self._canon = False
+        super(LongTurtleSerializer, self).__init__(store)
         self.keywords = {RDF.type: "a"}
         self.reset()
         self.stream = None
@@ -85,11 +82,34 @@ def addNamespace(self, prefix, namespace):
         super(LongTurtleSerializer, self).addNamespace(prefix, namespace)
         return prefix
 
+    def canonize(self):
+        """Apply canonicalization to the store.
+
+        This normalizes blank node identifiers and allows for deterministic
+        serialization of the graph.
+        """
+        if not self._canon:
+            return
+
+        namespace_manager = self.store.namespace_manager
+        store = to_canonical_graph(self.store)
+        content = store.serialize(format="application/n-triples")
+        lines = content.split("\n")
+        lines.sort()
+        graph = Graph()
+        graph.parse(
+            data="\n".join(lines), format="application/n-triples", skolemize=True
+        )
+        graph = graph.de_skolemize()
+        graph.namespace_manager = namespace_manager
+        self.store = graph
+
     def reset(self):
         super(LongTurtleSerializer, self).reset()
         self._shortNames = {}
         self._started = False
         self._ns_rewrite = {}
+        self.canonize()
 
     def serialize(
         self,
@@ -99,6 +119,7 @@ def serialize(
         spacious: Optional[bool] = None,
         **kwargs: Any,
     ) -> None:
+        self._canon = kwargs.get("canon", False)
         self.reset()
         self.stream = stream
         # if base is given here, use, if not and a base is set for the graph use that
diff --git a/test/test_serializers/test_serializer_longturtle.py b/test/test_serializers/test_serializer_longturtle.py
index c1761b6da..65821784e 100644
--- a/test/test_serializers/test_serializer_longturtle.py
+++ b/test/test_serializers/test_serializer_longturtle.py
@@ -167,7 +167,7 @@ def test_longturtle():
     g.bind("sdo", SDO)
 
     # run the long turtle serializer
-    output = g.serialize(format="longturtle")
+    output = g.serialize(format="longturtle", canon=True)
 
     # fix the target
     current_dir = Path.cwd()  # Get the current directory
diff --git a/test/test_serializers/test_serializer_longturtle_sort.py b/test/test_serializers/test_serializer_longturtle_sort.py
index fa192253d..044660e3e 100644
--- a/test/test_serializers/test_serializer_longturtle_sort.py
+++ b/test/test_serializers/test_serializer_longturtle_sort.py
@@ -62,7 +62,7 @@ def test_sort_semiblank_graph() -> None:
             graph.add((outer_node, EX.has, inner_node))
             graph.add((inner_node, RDFS.seeAlso, nested))
 
-        graph_text = graph.serialize(format="longturtle", sort=True)
+        graph_text = graph.serialize(format="longturtle", canon=True)
         if first_graph_text == "":
             first_graph_text = graph_text
 

From 88bf615c6aa28941ba77ec6bacff52c78da1da60 Mon Sep 17 00:00:00 2001
From: Edmond Chuc <edmond@kurrawong.ai>
Date: Wed, 3 Sep 2025 20:48:01 +1000
Subject: [PATCH 2/2] docs: fix docs build error by removing py obj reference
 to canon

---
 rdflib/plugins/serializers/longturtle.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/rdflib/plugins/serializers/longturtle.py b/rdflib/plugins/serializers/longturtle.py
index 1315181b1..2aaed36e6 100644
--- a/rdflib/plugins/serializers/longturtle.py
+++ b/rdflib/plugins/serializers/longturtle.py
@@ -41,7 +41,7 @@
 class LongTurtleSerializer(RecursiveSerializer):
     """LongTurtle, a Turtle serialization format.
 
-    When the optional parameter `canon` is set to True, the graph is canonicalized
+    When the optional parameter ``canon`` is set to :py:obj:`True`, the graph is canonicalized
     before serialization. This normalizes blank node identifiers and allows for
     deterministic serialization of the graph. Useful when consistent outputs are required.
     """