ShannonBase Javascript Engine

The image shows a discussion where someone complained after the release of MySQL 9.7. It can open the feature of supporting JS routine.

But don’t worry — ShannonBase already supports a JS engine, allowing you to write and execute JavaScript programs inside ShannonBase just like writing SQL stored procedures, and you can directly manipulate the local database within the JS program.

The choice to integrate a JavaScript (JS) engine is mainly to lower the development barrier, expand the database’s processing capabilities, and improve support for modern data formats (especially JSON). By embedding a lightweight JS engine (based on GraalVM) into the database, developers can write stored procedures and functions directly in JavaScript, rather than being limited to traditional SQL or stored procedure languages.

Expanding complex logic processing capabilities: SQL is a set-based declarative language, making it difficult to implement complex business logic or procedural processing. JavaScript, as an imperative programming language, is better suited for data transformations, custom algorithms, and other complex logic.

Bridging the “last mile” of JSON support: JSON data is naturally compatible with JavaScript. Using JavaScript to handle stored procedures allows developers to manipulate in-memory JSON data in a more native and efficient way.

Seamless compatibility with existing infrastructure: JavaScript stored programs work seamlessly with traditional SQL stored programs, InnoDB, Lakehouse, and the Rapid engine. This means you can write logic in JS while still leveraging Rapid’s high-performance computing capabilities.

Wider developer ecosystem: JavaScript is one of the most popular programming languages. Using JS allows a large number of frontend or backend developers to easily perform advanced development inside the database without learning complex SQL procedural languages.

More options: With the introduction of generative AI features, the JS engine provides flexibility for handling unstructured data and calling external APIs.

In summary, with the JS engine, ShannonBase achieves “processing logic where the data resides,” avoiding the latency of moving data out of the database just to handle logic, thus improving overall analysis and development efficiency.

Unlike Heatwave, ShannonBase does not use GraalVM as its JS engine; instead, it uses JerryScript. Compared to GraalVM, JerryScript offers the following advantages:

• Extremely lightweight and low resource usage: JerryScript is designed to run in less than 64 KB of RAM and 200 KB of ROM. For a database, this means the JS engine’s memory overhead can be kept minimal, leaving most memory resources for core tasks like data processing, query caching, and connection management. In contrast, GraalVM is designed for general-purpose applications and microservices, with a relatively large base runtime that significantly increases the database’s memory burden.
• Small code size, easy integration and maintenance: JerryScript is written in C99, and its compiled binary is very small (e.g., only about 258 KB when compiled for ARM Thumb-2). This makes it easy to embed into ShannonBase’s C++ codebase without significantly increasing the size of the installation package or binary files. Moreover, as a pure interpreter without complex JIT frameworks, its behavior and resource consumption are more predictable when executing user-provided JS code within the database process, making sandboxing easier.
• Fast startup and no JIT warm-up overhead: JerryScript has no JIT mechanism, so there is no “warm-up” process — code always starts being interpreted immediately. This is ideal for database scenarios, where stored procedures or functions are typically short and called frequently. JerryScript’s “zero startup cost” avoids CPU and memory spikes caused by JIT compilation, ensuring stable database performance and predictable response times.
• Mature embedded API and snapshot support: JerryScript provides a mature C API, making it easy for applications to call and embed directly. Its unique snapshot support allows JavaScript source code to be precompiled into bytecode. These snapshots can be preloaded in the database, further reducing runtime parsing and compilation overhead and significantly improving execution efficiency.

With this lightweight JS engine, ShannonBase now supports JavaScript stored functions. Since JerryScript itself, as a lightweight JS engine, does not include database connection APIs like ODBC or DAO in its core standard library, ShannonBase extends its capabilities by providing the sys.exec_sql interface to execute SQL queries within JS functions. Through sys.exec_sql, you can directly manipulate the database from inside a JS function. The sys.exec_sql function returns results in JSON format by default.

DELIMITER *
CREATE FUNCTION query_table() RETURNS TEXT
LANGUAGE JAVASCRIPT AS $$
function query() {
    return sys.exec_sql("select * from test");
}
return query();
$$*
DELIMITER ;

Additionally, exec_sql inherits all the states of the current connection.
Example execution:

mysql> DELIMITER *
mysql> CREATE FUNCTION query_table() RETURNS TEXT
    -> LANGUAGE JAVASCRIPT AS $$
    $> function query() {
    $>     return sys.exec_sql("select * from test");
    $> }
    $> return query();
    $> $$*
Query OK, 0 rows affected (0.01 sec)

mysql> DELIMITER ;
mysql> select query_table();
+---------------------------------------------------------------------------------------------------------------------------------------------+
| query_table()                                                                                                                               |
+---------------------------------------------------------------------------------------------------------------------------------------------+
| [{"score":1.1,"name":"n1","id":1,"gender":"m"},{"score":2.2,"name":"n2","id":2,"gender":"f"},{"score":3.3,"name":"n3","id":3,"gender":"m"}] |
+---------------------------------------------------------------------------------------------------------------------------------------------+
1 row in set (2.95 sec)

You can also use other features:

DELIMITER *
CREATE FUNCTION IS_EVEN (VAL INT) RETURNS INT
LANGUAGE JAVASCRIPT AS $$
function isEven(num) {
    return num % 2 == 0;
}
return isEven(VAL);
$$*
DELIMITER ;

If you like the project, please give me a star or submit a PR.

⭐ Star the repo
🧩 Submit PRs
🐞 Open Issues
💬 Join Discussion