Hacker News Reader

Comment by ekidd | Hacker News Reader

I've always wanted to like Go, but every time I get ~1,500 lines in a project, I remember my pain points. I totally see why other people like the current version of Go, but as it stands, it's not an ideal match for my brain.

Dependency management is a big pain point for me. I'm really glad to see several of my pain points on the list for this year, including another look at generics.

Generics are genuinely tricky: They allow you to write many kinds of useful functions in a type-safe manner, but every known approach for implementing them adds complexity to the language. C#, Java and Rust all bit the bullet and accepted (some) of this complexity. Maybe Go will find a sweet spot, preserving its simplicity but adding a bit of expressiveness?

Anyway, it pleases me to see that the Go team is thinking hard about this stuff. At the bare minimum, I'm going to be contributing code to other people's open source Go projects for the foreseeable future. :-)

View on HN

> C#, Java and Rust all bit the bullet and accepted (some) of this complexity. Maybe Go will find a sweet spot, preserving its simplicity but adding a bit of expressiveness?

That's exactly my hope. When i switched from Rust back to Go, i had a sigh of relief, i was able to prototype quickly and easily and my cognitive load felt much lower.

Strangely enough, this felt very similar to switching from NodeJS to Golang. In Node, i was constantly worried about what is async or sync and the dynamic nature of it made my code feel like the wild wild west. Both Rust and NodeJS put a lot of mental burden on me, in different ways - Go was definitely a sweet spot in both correctness and ease of use, and i hope they achieve that with Generics as well.

NOTE: My Rust programs felt vastly more secure than in go, and i miss that - that part was less cognitive load in favor of Rust. The struggle was mainly at the design phase, and i just wanted to mock up some code and types & borrowing posed many refactoring issues. I hope in the future strong Rust tooling will make refactoring a breeze.

> Generics are genuinely tricky: They allow you to write many kinds of useful functions in a type-safe manner, but every known approach for implementing them adds complexity to the language. C#, Java and Rust all bit the bullet and accepted (some) of this complexity. Maybe Go will find a sweet spot, preserving its simplicity but adding a bit of expressiveness?

I seriously doubt Go and the people who maintain it are going to do groundbreaking work in this area. It's and extremely developed area of language design (and still developing way ahead of where Go would ever go)

What "complexity" does simple parametric polymorphism (i.e. forall a) bring? The only thing I can think of is some extra syntax, which is far less complex than a codebase built upon a lack of parametricity. Hell, don't even allow user-defined parametric types and force everyone to stay with Go's parametric builtins but allow programmers to abstract over them. Seems like a no-brainer to me.

I think it is no surprise that one goal of Go is coding at large where large teams are involved. For single person projects that I think you are doing, many people want intellectually stimulating language where Go may fall short.

> I think it is no surprise that one goal of Go is coding at large where large teams are involved.

I don't buy that. There is no proof Go programming at large scales better than Java programming at large.

Go didn't reach the scale of Java programs yet. And no, Kub or Docker, while fairly large, are nothing compared to 15 y.o. multi-million line Java codebases. Go certainly needs less bureaucracy due to the ease of deployment, but it doesn't mean Go projects scale better in large teams.

> For single person projects that I think you are doing, many people want intellectually stimulating language where Go may fall short.

I really hate this kind of arguments. Features like generics aren't intellectually stimulating, they are here because people want to write type safe code. Context.Value(interface{})interface{} isn't type safe code.

Now tell me, what is more intellectually challenging? writing concurrent programs free of race conditions or generics?

> Go didn't reach the scale of Java programs yet. And no, Kub or Docker, while fairly large, are nothing compared to 15 y.o. multi-million line Java codebases.

I didn't claim Java projects do not scale. They do. I use Java all the time at my day job. I would use whatever makes sense. Maybe someday Rust is absolutely essential I would use it then.

> Features like generics aren't intellectually stimulating, they are here because people want to write type safe code.

When people need generics they can use languages with generics facility then. I am not arguing otherwise.

> When people need generics they can use languages with generics facility then. I am not arguing otherwise.

This is the point where Go's advocates disagree with its detractors. The latter tend to feel that you need generics (Just like you need to program in something higher level then assembly) far more frequently then they are told.

I want intellectually appealing code, not intellectually torturous.

Reply on news.ycombinator.com