Convert numerics to null-supported Pandas types (#679)

* Convert numerics to null-supported Pandas types

* update changelog

Author: michaelnchin

ChangeLog.md

@@ -11,6 +11,7 @@ Starting with v1.31.6, this file will contain a record of major features and upd
 - Enabled n-triples data for `%load` with Neptune Analytics ([PR #1](https://github.com/aws/graph-notebook/pull/671)) ( ([PR #2](https://github.com/aws/graph-notebook/pull/675)))
 - Removed unused options from `%load`([Link to PR](https://github.com/aws/graph-notebook/pull/662))
 - Made EncryptionKey optional in Neptune CloudFormation template ([Link to PR](https://github.com/aws/graph-notebook/pull/663))
+- Fixed unintended type coercion in results table with missing/null values ([Link to PR](https://github.com/aws/graph-notebook/pull/679))
 
 ## Release 4.5.1 (July 31, 2024)
 

src/graph_notebook/magics/graph_magic.py

@@ -241,7 +241,7 @@ def query_type_to_action(query_type):
 def oc_results_df(oc_res, oc_res_format: str = None):
     rows_and_columns = opencypher_get_rows_and_columns(oc_res, oc_res_format)
     if rows_and_columns:
-        results_df = pd.DataFrame(rows_and_columns['rows'])
+        results_df = pd.DataFrame(rows_and_columns['rows']).convert_dtypes()
         results_df = results_df.astype(str)
         results_df = results_df.map(lambda x: encode_html_chars(x))
         col_0_value = range(1, len(results_df) + 1)
@@ -893,7 +893,7 @@ def sparql(self, line='', cell='', local_ns: dict = None):
 
                         rows_and_columns = sparql_get_rows_and_columns(results)
                         if rows_and_columns is not None:
-                            results_df = pd.DataFrame(rows_and_columns['rows'])
+                            results_df = pd.DataFrame(rows_and_columns['rows']).convert_dtypes()
                             results_df = results_df.astype(str)
                             results_df = results_df.map(lambda x: encode_html_chars(x))
                             results_df.insert(0, "#", range(1, len(results_df) + 1))
@@ -1323,7 +1323,7 @@ def gremlin(self, line, cell, local_ns: dict = None):
                             query_res_deque.appendleft('x')
                             query_res = list(query_res_deque)
 
-                results_df = pd.DataFrame(query_res)
+                results_df = pd.DataFrame(query_res).convert_dtypes()
                 # Checking for created indices instead of the df itself here, as df.empty will still return True when
                 # only empty maps/lists are present in the data.
                 if not results_df.index.empty: