Fix errors of embedded table parsing (#13615)

Author: zhaozhiming

llama-index-core/llama_index/core/node_parser/relational/base_element.py

@@ -303,9 +303,14 @@ def get_nodes_from_elements(
 
                 # attempt to find start_char_idx for table
                 # raw table string regardless if perfect or not is stored in element.element
-                start_char_idx = ref_doc_text.find(str(element.element))
-                if start_char_idx >= 0:
-                    end_char_idx = start_char_idx + len(str(element.element))
+
+                if ref_doc_text:
+                    start_char_idx = ref_doc_text.find(str(element.element))
+                    if start_char_idx >= 0:
+                        end_char_idx = start_char_idx + len(str(element.element))
+                    else:
+                        start_char_idx = None
+                        end_char_idx = None
                 else:
                     start_char_idx = None
                     end_char_idx = None

llama-index-core/llama_index/core/node_parser/relational/llama_parse_json_element.py

@@ -33,7 +33,9 @@ def get_nodes_from_node(self, node: TextNode) -> List[BaseNode]:
         self.extract_table_summaries(table_elements)
         # convert into nodes
         # will return a list of Nodes and Index Nodes
-        return self.get_nodes_from_elements(elements, node)
+        return self.get_nodes_from_elements(
+            elements, node, ref_doc_text=node.get_content()
+        )
 
     def extract_elements(
         self,

llama-index-core/llama_index/core/node_parser/relational/unstructured_element.py

@@ -67,7 +67,9 @@ def get_nodes_from_node(self, node: TextNode) -> List[BaseNode]:
         self.extract_table_summaries(table_elements)
         # convert into nodes
         # will return a list of Nodes and Index Nodes
-        nodes = self.get_nodes_from_elements(elements, node)
+        nodes = self.get_nodes_from_elements(
+            elements, node, ref_doc_text=node.get_content()
+        )
 
         source_document = node.source_node or node.as_related_node_info()
         for n in nodes: