feat: Add Spark from_json function

velox/docs/functions/spark/json.rst

@@ -44,3 +44,31 @@ JSON Functions
         SELECT json_object_keys(''); -- NULL
         SELECT json_object_keys(1); -- NULL
         SELECT json_object_keys('"hello"'); -- NULL
+
+.. spark:function:: from_json(jsonString) -> array / map / row
+
+    Casts ``jsonString`` to an ARRAY, MAP, or ROW type, with the output type 
+    determined by the expression. Returns NULL, if the input string is unparsable.
+    Supported element types include BOOLEAN, TINYINT, SMALLINT, INTEGER, BIGINT, 
+    REAL, DOUBLE, VARCHAR, ARRAY, MAP and ROW. When casting to ARRAY or MAP, 
+    the element type of the array or the value type of the map must be one of 
+    these supported types, and for maps, the key type must be VARCHAR. Casting 
+    to ROW supports only JSON objects.
+    The current implementation has the following limitations.
+
+    * Does not support user provided options.
+
+    * Only supports partial result mode, which requires spark configuration spark.sql.json.enablePartialResults = true.
+
+    * Does not support single quotes as delimiters.  
+
+    * Does not support schemas that include a corrupt record column.  
+
+    Behaviors of the casts are shown with the examples below. ::
+
+        SELECT from_json('{"a": true}'); -- {'a'=true} // Output type: ROW({"a"}, {BOOLEAN()})
+        SELECT from_json('{"a": 1}'); -- {'a'=1} // Output type: ROW({"a"}, {INTEGER()})
+        SELECT from_json('{"a": 1.0}'); -- {'a'=1.0} // Output type: ROW({"a"}, {DOUBLE()})
+        SELECT from_json('["name", "age", "id"]'); -- ['name', 'age', 'id'] // Output type: ARRAY(VARCHAR())
+        SELECT from_json('{"a": 1, "b": 2}'); -- {'a'=1, 'b'=2} // Output type: MAP(VARCHAR(),INTEGER())
+        SELECT from_json('{"a": {"b": 1}}'); -- {'a'={b=1}} // Output type: ROW({"a"}, {ROW({"b"}, {INTEGER()})})

velox/functions/sparksql/registration/RegisterSpecialForm.cpp

@@ -18,6 +18,7 @@
 #include "velox/expression/SpecialFormRegistry.h"
 #include "velox/functions/sparksql/specialforms/AtLeastNNonNulls.h"
 #include "velox/functions/sparksql/specialforms/DecimalRound.h"
+#include "velox/functions/sparksql/specialforms/FromJson.h"
 #include "velox/functions/sparksql/specialforms/MakeDecimal.h"
 #include "velox/functions/sparksql/specialforms/SparkCastExpr.h"
 
@@ -44,6 +45,9 @@ void registerSpecialFormGeneralFunctions(const std::string& prefix) {
       "cast", std::make_unique<SparkCastCallToSpecialForm>());
   registerFunctionCallToSpecialForm(
       "try_cast", std::make_unique<SparkTryCastCallToSpecialForm>());
+  exec::registerFunctionCallToSpecialForm(
+      FromJsonCallToSpecialForm::kFromJson,
+      std::make_unique<FromJsonCallToSpecialForm>());
 }
 } // namespace sparksql
 } // namespace facebook::velox::functions

velox/functions/sparksql/specialforms/CMakeLists.txt

@@ -16,9 +16,10 @@ velox_add_library(
   velox_functions_spark_specialforms
   AtLeastNNonNulls.cpp
   DecimalRound.cpp
+  FromJson.cpp
   MakeDecimal.cpp
   SparkCastExpr.cpp
   SparkCastHooks.cpp)
 
 velox_link_libraries(velox_functions_spark_specialforms fmt::fmt
-                     velox_expression)
+                     velox_functions_json velox_expression)