TagNode.merge_text_nodes: Also drops empty text nodes

Optimizations are applied as well.

_delb/nodes.py

@@ -2258,9 +2258,20 @@ def location_path(self) -> str:
     def merge_text_nodes(self):
         """
         Merges all consecutive text nodes in the subtree into one.
+        Text nodes without content are dropped.
         """
-        for node in self.iterate_descendants(is_text_node):
-            node._merge_appended_text_nodes()
+        with _wrapper_cache:
+            empty_nodes: list[TextNode] = []
+
+            for node in self.iterate_descendants():
+                if not isinstance(node, TextNode):
+                    continue
+                node._merge_appended_text_nodes()
+                if not node.content:
+                    empty_nodes.append(node)
+
+            for node in empty_nodes:
+                node.detach()
 
     @property
     def namespace(self) -> Optional[str]:

tests/test_tag_node.py

@@ -460,6 +460,16 @@ def test_make_node_in_context_with_namespace():
     assert node._etree_obj.tag == "{https://name.space}foo"
 
 
+@pytest.mark.parametrize(
+    ("child_nodes", "expected_count"),
+    (([""], 0), ([" "], 1), ([" ", " "], 1), (["", "", tag("child"), "", ""], 1)),
+)
+def test_merge_text_nodes(child_nodes, expected_count):
+    node = new_tag_node("node", children=child_nodes)
+    node.merge_text_nodes()
+    assert len(node) == expected_count
+
+
 def test_names(sample_document):
     root = sample_document.root