Show HN: Java REST without annotations, DI nor reactive streams (github.com)

The worst parts of Spring is Spring. Nest lacks the IDE, the performance in the JVM, etc... there is nothing to gain.

First question: is there maven already available for this library? Mostly to ease upgrades in the future.

Second question: what strategy would you recommend for adding https on the communication channel?

Thanks.

https://sparkjava.com/

You instantiate a RestApi object and then . . . . magic?

    public class Foo {
      public Foo(Bar bar, Baz baz) {}
    }

Oh well, will continue using java far away from that framework.

Aka using a library

I'm now working on a Java side project, and while typically in the past I've just done the backend in java and then used rails for the web UI and had them share a DB, I'm trying to see if I can use Java for the web part without going insane.

Part of the problem is that historically the big players in java web are HUGE enterprises that was to be able to have 50 teams all do a small part of a backend in parallel and then just deploy them all together. Thus was born the servlet API and application servers.

But there's so many assumptions and bizarre requirements that come out of trying to do this perverse form of engineering that the whole thing ends up nigh unusable for someone who could otherwise just "rails g" 85% of their project.

Jetty has always struck me as a bit of a middle man where if you want, you can do the servlet thing but it's also a production grade application framework that doesn't force you to do the java EE dance if you don't want to. Though there is some leakage. But it's a lot less magical than Spring and is also just the http server bits, not the rest of the db and view and etc.

But since virtual threads are pretty stable now, I really want to use them, and jetty is the first reasonably complete and robust option that seems to have included support for them.

So - you know how these things go. I'm currently writing an HTTP url path router that supports the rails syntax from routes.rb. And then I'm going to write a Handler implementation that wrangles all the database stuff and does convenient/terse parsing of params (like how rails folds path params, url params, and form params into a single params object) and rendering of responses (so I can render a json object without having to call Content.Sink.[...] and use gson all over the place.

It's meant to all be very non magical and you can step through the code in a debugger and not see a billion reflective invocations of methods. And I'm hoping I can make the API of my Handler convenient enough that you don't regret that it's not annotation magic-based.

Also - I know there are various attempts at easier/less annoying Java web things like vertx and such, but they a) don't support virtual threads yet, and b) many of them are small enough im worried they'll rot eventually. Jetty meanwhile isn't going anywhere.

Probably worth noting that this depends entirely on the annotation - they can (and many do) run at compile time, and can provide very strong safety guarantees.

Many (most? all? I dunno) of the bloated server-side DI frameworks do not do this though, and I 100% agreed that it's can be a truly awful experience.

However, there are some real issues with annotations in Spring/Java: - Application will sometimes run just fine without annotation processor/interceptor. Think of `@EnableScheduling` in Spring: you won't know that `@Scheduled` is not working (because of missing `@EnableScheduling`) until you observe that method is not executed. In this case static code is a clear win. - Annotation order: not all annotation processors/interceptors in Spring support specifying order. Annotation order in the code doesn't matter: it is lost during compilation. Good luck figuring out what is applied first in a method with `@Retry`, `@Transactional` and `@Cached` - will retry be executed within transaction or each retry will have its own transaction? This also is easily solved with static code instead of annotations.

As for compile-time error vs runtime-error: personally I don't really care as long as there is any error (which is not always the case in the first example) during the build/test/init/assembly phase. When I'm writing SQL queries in the code, I'm getting SQL parsing/compilation errors during application runtime - but that's fine, because I've written SQL-s against DB execution engine. When I'm writing Spark SQL job, I'm getting errors during query planning phase - and that's also fine, because I'm writing code against Spark's execution engine. Writing annotations against "annotation execution engine" (annotation processor/interceptors) doesn't seem any different or wrong in principle. Although, there are things that could be improved.

Stacktraces: there are a few additional interceptor method calls in the stacktrace when annotations are in use, however, most of the complexity comes from library/framework structure and developer's familiarity with it. Spring covers a lot of use cases thus it has its share of complexity. I'm not sure if "Spring without annotations" would be noticeably easier to work with, although I assume that feature-parity with Spring (MVC) is not a goal of this project so it probably will be easier to understand.

Much better it fails at compile time.

However, Groovy is also a great scripting language with incredibly powerful runtime meta-programming features. In short, that means you're stuck inside often stripped down (Jenkins) or ill-conceived DSLs that you either know by heart, know someone who knows it, or are in for a world of hurt trying to do basically anything.

With Kotlin DSL - even though the stack got even more complex, including embedded scripting host inside embedded JVM and all that - you get auto-completion in the IDE and "go to definition". That resolves half the problem with Groovy, which is discoverability. The other half - the byzantine object model and PERL-like "there's more than 1 way to do it" - won't go away just by changing the scripting language. Still, it's better.

There's also the thing about versions and compatibility between Gradle versions, Kotlin versions, and plugins versions. Finding the right combination takes so much effort that seemingly nobody ever bother to update the build tools in projects. Unless there's a "build engineer", which sounds kind of dystopic, but after working with Gradle for half a year I have to admit that managing the builds properly is indeed a full-time job.

> Hardware is cheap these days, developer time is more expensive. Use Node or Python.

It's not that simple. Sometimes, latency matters. Sometimes you really need something to happen in 20ms. But then you won't be using Java, or what pjmpl would say, you wouldn't use the stock JVM, but something that provides AOT compilation.

There are so many dimensions you need to consider when choosing an implementation language for something, and Java can be an optimal choice in many situations. As unfortunate as I think it is, the problem is not technical, but cultural. People who started programming in Java will have this rigid, rule-based concept of what you can and cannot do. Their minds are semi-permanently stuck in Java-like mold. Since Java is so all-encompassing, they are never confronted on their beliefs. Echo chamber. Cargo culting. Prevalent in all monocultures. The problem is when they switch to another language and infect it with beliefs that stopped being well founded due to the changed circumstances. They do their best to fit the new language into Java-esque shape, no matter how pointless it seems.

Kotlin is the biggest victim of this. There's so clearly visible divide between "make Java great again" crowd vs. "it's actually a nice language, why not use it for what it is?" crowd. They produce drastically different libraries, have very different goals, and do very different projects. Different values, methods, and outcomes. In itself, it's maybe OK, but... try being stuck in the wrong camp... and stay sane.

        public SomeService() {
           Spark.post("some/endpoint", this::someEndpoint, gson::toJson);
           Spark.get("another/endpoint/:id". this::anotherEndpoint, gson::toJson);
           //...
         }


        private ResponseType someEndpoint(Request request, Response response) {
           Something some = gson.fromJson(request.body(), Something.class);

           return new ResponseType(some.a, logic(some));
        }

        private ResponseType2 anotherEndpoint(Request request, Response response) {
           return new ResponseType2(otherLogic(request.param("id"));
        }

Not to detract from your project, of course.

Not really. Spring has always had problems with pointless complexity and turning what should be compile time errors into runtime ones. Its original schtick was DI, which you've always been able to do by just... sending dependencies in through the constructor, lol? No framework or magic needed. The emperor really has no clothes and I wish more people would admit that.

With an annotation-based framework, how do you do the same? You could find usages of the annotations, and then manually read the thousands of places “@GET” is referenced?

Imagine I'm looking at a class that is annotated. What is the next step to find the code that gets executed for the annotations?

If you inject the bean and call the method you will get caching (because you are using the proxy). If you call the method from within the bean itself however you're not using the proxy and you won't get caching.

It's stuff like this on steroids when you start mixing annotations that makes it really difficult to reason about the code.

From my exprience there are multiple reasons - The main one: I've learned tons from reading code. The ability to jump into code that does X to see how it's implemented is worth a lot to me. This is something I love about the jvm ecosystem which is unfortunately not present in many other languages. - Many libraries/frameworks (Spring Data comes to mind) provide very leaky abstractions. Understanding how the underlying technology is used can make a big difference. - Address questions when the documentation is not clear enough. - And last: to discover bugs in libraries. This is not common, but it happens.

REST APIs do. SOAP has the same "feature" of failing on unknown properties and it made API evolution very hard.

the point of a framework is to hide some of that from you. The point of abstraction is that you _should not_ have to know how the server mounts the handlers, just that it does and it does it via a contract specified in the documentation.

These are problems that every web application has to solve. You can use an existing framework or you can write your own. It might seem like there's a "hand-wire everything" option but in any mature system that turns into "write your own framework" as developers get tired of typing the same boilerplate over and over.

For example, I can tell you right now that the team is going to get tired of typing `requestCycle.parseBody()` at the top of almost every method. Some clever sod will eventually figure out how to use AOP to make it implicit. Just like JAX-RS.

Let's say we add some server-side tests that verify that the method is POST and the path is /make-greeting. Have we now verified that the endpoint is correct? No - because the client was in fact posting to /create-greeting.

We know the @POST and @Path annotations work; they are covered by the JAX-RS implementation tests. We can look at those annotations and know that the endpoint is a POST to /make-greeting. Testing that brings minimal value.

The test we need is something that ensures the client and server both agree on the contract. By definition that can't be a server-side test, it's an e2e test. The good news is that it doesn't have to be a complicated e2e test; it just needs to ensure the method is called once.

Maybe you're publishing a pure API? Add swagger, auto-generate the openapi spec, and diff against that. This works especially well if openapi is how you're publishing your documentation.

> and the body parser

If you're really concerned, it's pretty easy to write explicit tests for JSON serialization/deserialization. The good news is that you only need one test (or two for bidirectional) for each class. In both large and small companies I've never found this necessary.

reading the spring documentation usually help :)

https://github.com/minebreaker/Abukuma/tree/master/documenta...

It's a half-baked project and may never be finished.