sync data/cdata children when setting element value (fixes #6)

With parse_cdata=True the document is parsed with data nodes, so the
element text lives in node_data/node_cdata children which take
precedence over the element's own value during printing. Setting
node.value only updated the element value, leaving serialization
unchanged. Now the first data/cdata child is updated as well, and any
extra data/cdata children are removed.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: Marc-Etienne Barrut

rapidxml/c_ext/src/base_object.cpp

@@ -163,7 +163,35 @@ static int rapidxml_BaseObject_setvalue(rapidxml_BaseObject* self,
   if (!PyArg_Parse(arg, "s*", &value)) {
     return -1;
   }
-  self->underlying_obj->value(self->document->allocate_string(static_cast<const char*>(value.buf)));
+  char* new_value = self->document->allocate_string(static_cast<const char*>(value.buf));
+  self->underlying_obj->value(new_value);
+  if (IS_NODE(reinterpret_cast<PyObject*>(self))) {
+    rapidxml::xml_node<>* node =
+      static_cast<rapidxml::xml_node<>*>(self->underlying_obj);
+    if (node->type() == rapidxml::node_element) {
+      /*
+      ** When the document is parsed with data nodes (parse_cdata=True),
+      ** the element value is held by data/cdata children, which take
+      ** precedence over the element value when printing.
+      ** Keep them in sync: update the first one, drop the others.
+      */
+      rapidxml::xml_node<>* child = node->first_node();
+      bool found = false;
+      while (child != NULL) {
+        rapidxml::xml_node<>* next = child->next_sibling();
+        if (child->type() == rapidxml::node_data ||
+            child->type() == rapidxml::node_cdata) {
+          if (!found) {
+            child->value(new_value);
+            found = true;
+          } else {
+            node->remove_node(child);
+          }
+        }
+        child = next;
+      }
+    }
+  }
   PyBuffer_Release(&value);
   return 0;
 }

tests/test_issue6.py

@@ -0,0 +1,32 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+#
+# File: test_issue6.py
+# by Arzaroth Lekva
+# arzaroth@arzaroth.com
+#
+
+import rapidxml
+
+def test_set_value_with_parse_cdata():
+    r = rapidxml.RapidXml(b'<root><loc>old value</loc></root>',
+                          parse_cdata=True)
+    loc = r.first_node().first_node("loc")
+    loc.value = "new value"
+    assert loc.value == "new value"
+    assert r.unparse() == '<root><loc>new value</loc></root>'
+
+def test_set_value_with_cdata_node():
+    r = rapidxml.RapidXml(b'<root><loc><![CDATA[old value]]></loc></root>',
+                          parse_cdata=True)
+    loc = r.first_node().first_node("loc")
+    loc.value = "new value"
+    assert loc.value == "new value"
+    assert r.unparse() == '<root><loc><![CDATA[new value]]></loc></root>'
+
+def test_set_value_without_parse_cdata():
+    r = rapidxml.RapidXml(b'<root><loc>old value</loc></root>')
+    loc = r.first_node().first_node("loc")
+    loc.value = "new value"
+    assert loc.value == "new value"
+    assert r.unparse() == '<root><loc>new value</loc></root>'