Comment by fetchfetch | Hacker News Reader

Comment by fetchfetch | Hacker News Reader

Parent Full thread

After having coded tens of thousands of lines in Golang for 4 years, there are a list of probably hundreds of grievances I have about the language itself. The problem is that they're generally outweighed by the simplicity, consistency, and clarity of the language. Is the error system terrible? Obviously. Lack of abstract data types? Awful. Is Rust one thousand fold safer and more syntactically powerful by comparison? Sure.

But the fact is, in a world where we as developers often are learning and working in a new language every year or two, time-to-productivity and proficiency in a given language is one of the most important, if not the most important, metrics for a language. And Golang knocks this out of the park with its obvious syntax and great tooling. You can bring in a new junior programmer onto your Golang stack and have then turning out useful pull requests in a week or two, which I can't say for languages like Rust.

I _hate_ much of Golang's engineering, but for collaborative programming I think anyone would agree that it's a highly productive language.

Why learn a new language just to learn a new language for production stuff? I looked at at switching to Go from Java for my flagship product. I couldn't find enough compelling reasons to switch. It didn't offer enough benefits to outweigh the benefits of Java + Spring Boot. As a result, Java it is.

Now for a new product that I might sell as traditional installed software, Go makes some sense here since it simplifies install. My clients wouldn't need to install and maintain the VM.

In the end I wonder why switch to a new stack just because? I understand learning the languages to see if they are better tools. I understand learning for the sake of learning. But why bet a business on a new language just because it's cool?

My team just chose golang instead of java or dotnet core for a new crud app. I'm not especially fond of go, but mean length of feedback loops are worth it (so far). We have a bunch of tests, and they all run in a few hundredths of a second. Same coverage in java is usually several seconds at least. Often up to 30s if running tests with maven.

Standing up the app is also instantaneous. Spring boot apps with hibernate often take at least ten seconds. In prod this doesn't matter. But a cheap app start means integration tests are cheap. An fs hook runs both unit and integration tests every time the code changes.

ps. all the things people hate about go are true

There are lighter options in java world, my dropwizard server with JDBI starts in 1.2sec.

I'm genuinely unsure if this is a joke. If you're running that on a modern computer, your server is spending billions of clock cycles starting up. What on Earth is it doing? :)

If I write a Go server, I'm unhappy if it takes more than 0.2 seconds to start.

Depends what your server is doing, java app binded to http port will take probably same amount of time.

My dropwizard server contains almost all possible batteries: ssl, logging, http filters, static content, metrics, DB connection pooling, static content serving, all kinds of URL parameter parsing and URL routing, DI framework, ORM, thread pooling. I think it just inspects and initializes all this data structures in memory, load libraries and classes, etc.

> My clients wouldn't need to install and maintain the VM

Why should they?

One can always deliver everything together.

Not much different than static linking.

Also there are AOT compilers and with Java 9 we are getting a linker to create customized JREs.

I'm waiting for the 9 features. That should make things easier. Since I'm targeting small physician practices, I need to make everything as simple as possible. I've even toyed with Raspberry Pi as the install device. From what I've read, Go is a good option for this.

Reply on news.ycombinator.com