feat: add native support for get_json_object expression (#3747)

Author: andygrove

docs/source/user-guide/latest/expressions.md

@@ -90,6 +90,12 @@ Expressions that are not Spark-compatible will fall back to Spark by default and
 | Substring       | Yes               |                                                                                                            |
 | Upper           | No                | Results can vary depending on locale and character set. Requires `spark.comet.caseConversion.enabled=true` |
 
+## JSON Functions
+
+| Expression    | Spark-Compatible? | Compatibility Notes                                                                           |
+| ------------- | ----------------- | --------------------------------------------------------------------------------------------- |
+| GetJsonObject | No                | Spark allows single-quoted JSON and unescaped control characters which Comet does not support |
+
 ## Date/Time Functions
 
 | Expression     | SQL                          | Spark-Compatible? | Compatibility Notes                                                                                                              |

docs/spark_expressions_support.md

@@ -245,7 +245,7 @@
 ### json_funcs
 
 - [ ] from_json
-- [ ] get_json_object
+- [x] get_json_object
 - [ ] json_array_length
 - [ ] json_object_keys
 - [ ] json_tuple

native/Cargo.lock

@@ -33,7 +33,8 @@ datafusion = { workspace = true }
 chrono-tz = { workspace = true }
 num = { workspace = true }
 regex = { workspace = true }
-serde_json = "1.0"
+# preserve_order: needed for get_json_object to match Spark's JSON key ordering
+serde_json = { version = "1.0", features = ["preserve_order"] }
 datafusion-comet-common = { workspace = true }
 futures = { workspace = true }
 twox-hash = "2.1.2"

native/spark-expr/Cargo.toml

@@ -186,6 +186,10 @@ pub fn create_comet_physical_fun_with_eval_mode(
             let func = Arc::new(crate::string_funcs::spark_split);
             make_comet_scalar_udf!("split", func, without data_type)
         }
+        "get_json_object" => {
+            let func = Arc::new(crate::string_funcs::spark_get_json_object);
+            make_comet_scalar_udf!("get_json_object", func, without data_type)
+        }
         _ => registry.udf(fun_name).map_err(|e| {
             DataFusionError::Execution(format!(
                 "Function {fun_name} not found in the registry: {e}",