Improve performance of json functions

Avoid allocating heap ByteBuffer used by InputStreamReader.

Author: Dith3r

core/trino-main/src/main/java/io/trino/util/JsonUtil.java

@@ -63,6 +63,7 @@
 import java.io.IOException;
 import java.io.InputStreamReader;
 import java.io.OutputStream;
+import java.io.StringReader;
 import java.math.BigDecimal;
 import java.util.ArrayList;
 import java.util.HashMap;
@@ -105,6 +106,9 @@
 
 public final class JsonUtil
 {
+    // StringReader outperforms InputStreamReader for small inputs. Limit based on Jackson benchmarks {@link https://github.com/FasterXML/jackson-benchmarks/pull/9}
+    private static final int STRING_READER_LENGTH_LIMIT = 8192;
+
     private JsonUtil() {}
 
     // This object mapper is constructed without .configure(ORDER_MAP_ENTRIES_BY_KEYS, true) because
@@ -124,7 +128,12 @@ public static JsonParser createJsonParser(JsonFactory factory, Slice json)
             throws IOException
     {
         // Jackson tries to detect the character encoding automatically when using InputStream
-        // so we pass an InputStreamReader instead.
+        // so we pass StringReader or an InputStreamReader instead.
+        if (json.length() < STRING_READER_LENGTH_LIMIT) {
+            // StringReader is more performant than InputStreamReader for small inputs
+            return factory.createParser(new StringReader(new String(json.getBytes(), UTF_8)));
+        }
+
         return factory.createParser(new InputStreamReader(json.getInput(), UTF_8));
     }