Update

Author: philo-he

velox/core/QueryConfig.h

@@ -370,7 +370,8 @@ class QueryConfig {
       "presto.array_agg.ignore_nulls";
 
   /// If true, Spark function's behavior is ANSI-compliant, e.g. throws runtime
-  /// exception instead of returning null on invalid inputs.
+  /// exception instead of returning null on invalid inputs. It affects only
+  /// functions explicitly marked as "ANSI compliant".
   /// Note: This feature is still under development to achieve full ANSI
   /// compliance. Users can refer to the Spark function documentation to verify
   /// the current support status of a specific function.

velox/docs/configs.rst

@@ -1000,7 +1000,7 @@ Spark-specific Configuration
      - bool
      - false
      - If true, Spark function's behavior is ANSI-compliant, e.g. throws runtime exception instead
-       of returning null on invalid inputs.
+       of returning null on invalid inputs. It affects only functions explicitly marked as "ANSI compliant".
        Note: This feature is still under development to achieve full ANSI compliance. Users can
        refer to the Spark function documentation to verify the current support status of a specific
        function.

velox/docs/spark_functions.rst

@@ -2,10 +2,11 @@
 Spark Functions
 ***********************
 
-The semantics of Spark functions align with Spark 3.5. In the function descriptions,
-a function is marked as *ANSI compliant* if it adheres to ANSI standard, subject to
-the ``spark.ansi_enabled`` configuration. Otherwise, it simply follows Spark's semantics
-in ANSI OFF mode.
+The semantics of Spark functions align with
+`Spark 3.5 <https://spark.apache.org/docs/3.5.0/api/sql/index.html>`_.
+In the function descriptions, a function is marked as *ANSI compliant*
+if it adheres to ANSI standard, subject to the :doc:`spark.ansi_enabled <configs>`
+configuration. Otherwise, it simply follows Spark's semantics in ANSI OFF mode.
 
 .. toctree::
     :maxdepth: 1