upgrade simdjson third party to 4.1.0 (facebookincubator#15306)

Summary:
Pull Request resolved: facebookincubator#15306

allow-large-files

This commit was generated using `mgt import`.
pristine code for third-party libraries:
third-party/simdjson

uuid_cc2796e6e0604da2800cc1275df8ef85

update simdjson requested as we were running into ubsan errors in aarch64. i have validated that they go away with this change.

Reviewed By: kgpai

Differential Revision: D85209914

fbshipit-source-id: c2972c81a57452f71e98d79d86466b809a160c61

Author: kevyang

CMake/resolve_dependency_modules/README.md

@@ -30,7 +30,7 @@ by Velox. See details on bundling below.
 | xsimd             | 10.0.0          | Yes      ||
 | re2               | 2024-07-02      | Yes      ||
 | fmt               | 11.2.0          | Yes      | Used API must be fmt 9 compatible |
-| simdjson          | 3.13.0          | Yes      ||
+| simdjson          | 4.1.0           | Yes      ||
 | faiss             | 1.11.0          | Yes      ||
 | folly             | v2025.04.28.00  | Yes      ||
 | fizz              | v2025.04.28.00  | No       ||

CMake/resolve_dependency_modules/simdjson.cmake

@@ -13,10 +13,10 @@
 # limitations under the License.
 include_guard(GLOBAL)
 
-set(VELOX_SIMDJSON_VERSION 3.13.0)
+set(VELOX_SIMDJSON_VERSION 4.1.0)
 set(
   VELOX_SIMDJSON_BUILD_SHA256_CHECKSUM
-  07a1bb3587aac18fd6a10a83fe4ab09f1100ab39f0cb73baea1317826b9f9e0d
+  78115e37b2e88ec63e6ae20bb148063a9112c55bcd71404c8572078fd8a6ac3e
 )
 set(
   VELOX_SIMDJSON_SOURCE_URL

CMakeLists.txt

@@ -616,7 +616,7 @@ if(${VELOX_BUILD_MINIMAL_WITH_DWIO} OR ${VELOX_ENABLE_HIVE_CONNECTOR} OR VELOX_E
 endif()
 
 velox_set_source(simdjson)
-velox_resolve_dependency(simdjson 3.13.0)
+velox_resolve_dependency(simdjson 4.1.0)
 
 velox_set_source(FastFloat)
 velox_resolve_dependency(FastFloat)

scripts/setup-versions.sh

@@ -26,7 +26,7 @@ ARROW_VERSION="15.0.0"
 DUCKDB_VERSION="v0.8.1"
 PROTOBUF_VERSION="21.8"
 XSIMD_VERSION="10.0.0"
-SIMDJSON_VERSION="3.13.0"
+SIMDJSON_VERSION="4.1.0"
 CPR_VERSION="1.10.5"
 DOUBLE_CONVERSION_VERSION="v3.1.5"
 RANGE_V3_VERSION="0.12.0"

velox/functions/prestosql/JsonFunctions.cpp

@@ -33,6 +33,12 @@ template <typename T>
 static simdjson::error_code validate(T value) {
   SIMDJSON_ASSIGN_OR_RAISE(auto type, value.type());
   switch (type) {
+    case simdjson::ondemand::json_type::unknown: {
+      // In simdjson 4.0.7+, unknown type is returned for invalid JSON values
+      // like NaN, Infinity, etc. Return INCORRECT_TYPE to indicate invalid
+      // JSON.
+      return simdjson::INCORRECT_TYPE;
+    }
     case simdjson::ondemand::json_type::array: {
       SIMDJSON_ASSIGN_OR_RAISE(auto array, value.get_array());
       for (auto elementOrError : array) {
@@ -389,6 +395,12 @@ class JsonParseImpl {
   simdjson::error_code generateViews(T value) const {
     SIMDJSON_ASSIGN_OR_RAISE(auto type, value.type());
     switch (type) {
+      case simdjson::ondemand::json_type::unknown: {
+        // In simdjson 4.0.7+, unknown type is returned for invalid JSON values
+        // like NaN, Infinity, etc. Return INCORRECT_TYPE to indicate invalid
+        // JSON.
+        return simdjson::INCORRECT_TYPE;
+      }
       case simdjson::ondemand::json_type::array: {
         SIMDJSON_ASSIGN_OR_RAISE(auto array, value.get_array());
         return generateViewsFromArray<kNeedNormalize>(array);
@@ -761,6 +773,12 @@ struct JsonExtractImpl {
       // contents directly) and we might miss invalid JSON.
       SIMDJSON_ASSIGN_OR_RAISE(auto vtype, v.type());
       switch (vtype) {
+        case simdjson::ondemand::json_type::unknown: {
+          // In simdjson 4.0.7+, unknown type is returned for invalid JSON
+          // values like NaN, Infinity, etc. Return INCORRECT_TYPE to indicate
+          // invalid JSON.
+          return simdjson::INCORRECT_TYPE;
+        }
         case simdjson::ondemand::json_type::object: {
           SIMDJSON_ASSIGN_OR_RAISE(
               auto jsonStr, simdjson::to_json_string(v.get_object()));

velox/functions/prestosql/JsonFunctions.h

@@ -303,6 +303,9 @@ struct JsonSizeFunction {
           case simdjson::ondemand::json_type::null:
             singleResultSize = 0;
             break;
+          case simdjson::ondemand::json_type::unknown: {
+            return simdjson::INCORRECT_TYPE;
+          }
         }
       }
       return simdjson::SUCCESS;

velox/functions/prestosql/benchmarks/JsonExprBenchmark.cpp

@@ -243,7 +243,7 @@ struct SimdjsonParseFunction {
     } else {
       SIMDJSON_ASSIGN_OR_RAISE(auto doc, simdjsonParse(paddedJson));
       // Parse paddedJson again.
-      simdjsonParse(paddedJson).get(jsonDoc);
+      std::ignore = simdjsonParse(paddedJson).get(jsonDoc);
     }
     result = true;
     return simdjson::SUCCESS;

velox/functions/prestosql/tests/JsonCastTest.cpp

@@ -1122,12 +1122,12 @@ TEST_F(JsonCastTest, toDouble) {
       JSON(),
       DOUBLE(),
       {"Infinity"_sv},
-      "The JSON document has an improper structure");
+      "The JSON element does not have the requested type");
   testThrow<JsonNativeType>(
       JSON(),
       DOUBLE(),
       {"NaN"_sv},
-      "The JSON document has an improper structure");
+      "The JSON element does not have the requested type");
 
   testThrow<JsonNativeType>(
       JSON(),

velox/functions/prestosql/tests/JsonFunctionsTest.cpp

@@ -310,7 +310,8 @@ TEST_F(JsonFunctionsTest, jsonParse) {
       R"([{"response":"[\"fusil a peinture\",\"�uD83E\\uDE2Dau bois\"]"}])");
 
   VELOX_ASSERT_THROW(
-      jsonParse(R"({"k1":})"), "The JSON document has an improper structure");
+      jsonParse(R"({"k1":})"),
+      "The JSON element does not have the requested type");
   VELOX_ASSERT_THROW(
       jsonParse(R"({:"k1"})"), "The JSON document has an improper structure");
   // The exact exception message can change based on the simdjson version used.