Fork me on GitHub

SparkJava and Javalin comparison

People often ask about the differences between Spark and Javalin. Javalin originally started as a fork of Spark, and only covered a subset of Spark’s functionality. The project became popular very fast, and reached feature parity with Spark within months.

Javalin now offers a lot of features that are not available in Spark:

Syntax differences

Javalin has the concept of a Handler. A Handler is void and takes a Context, which wraps HttpServletRequest and HttpServletResponse. You operate on both the request and response through this Context.

javalin.get("/path", ctx -> ctx.result("Hello, World!"));
javalin.after("/path", ctx -> ctx.result("Actually, nevermind..."));

Spark on the other hand has Routes and Filters . Both Route and Filter in Spark take (Request, Response) as input. Route has a return value (Object), while Filter is void.

Request and Response wrap HttpServletRequest and HttpServletResponse, respectively.

Spark.get("/path", (req, res) -> "Hello, World!");
Spark.after("/path", (req, res) -> res.body("Actually, nevermind..."));

This means that in Javalin you always call ctx.result(string) when you want to update the result, whereas in Spark you have to return string in Routes and use response.body(string) in Filters.

Javalin’s approach leads to a much more consistent API, both for the previous and the next examples:


javalin.get("/", ctx -> ctx.redirect("/new-path"));
// vs
Spark.get("/", (req, res) -> {
    res.redirect("/new-path"); // can't return here, the redirect method is void
    return ""; // if you return null here you get a 404

JSON mapping

javalin.get("/", ctx -> ctx.json(object));
// vs
Spark.get("/", (req, res) -> object, new JsonTransformer());


javalin.get("/", ctx -> ctx.render("path", model));
// vs
Spark.get("/", (req, res) -> new ModelAndView(model, "path"), new TemplateEngine());

Code base and performance

Javalin’s code-base is a lot smaller than Spark’s (about one third the size) and written in Kotlin. All methods have proper nullable types. There are hundreds of tests (there is more test code than actual code). Javalin is about twice as fast, according to the TechEmpower Benchmarks.

Do you like Javalin?