[japano1se]

Don't move the goalposts. Global state is not the same issue as immutable data. And even if it was, Go has the `const` keyword anyway, making it entirely irrelevant to the OP article.

This fad of using immutability everywhere and treating it like a silver bullet absolutely is new and is a fad.

[fauigerzigerk]

The article isn't advocating "immutability everywhere". It complains about Go making it difficult to create immutable (or otherwise customized) data structures at all. So it is you who is moving the goalposts.

I mentioned global mutable state as an anecdote and also as an example on one extreme end of the spectrum that should make it clear what the problem is in principle: Not knowing what code changes what state.

But I see that you have decided to avoid debating the core issue entirely.

Your mention of Go's const keyword in this context leaves me scratching my head as you probably know that it doesn't allow you to define immutable maps, global or otherwise.

[japano1se]

Thanks for the "no u".

One thing you're correct about: I have avoided debating it, because debating concepts like this is pointless and has nothing to do with real-world code. Which brings me back to the point I made on the article: why does he care so much that he can't have an immutable map? What problem can he not solve as a result of it? The example he's given is a toy problem, and the constraints that make him want such a thing are left unspecified.

[fauigerzigerk]

>I have avoided debating it, because debating concepts like this is pointless and has nothing to do with real-world code.

It does have everything to do with my real world code. There is not a single piece of code I have ever written that let me get away with not keeping track of all the mutations that could possibly occur.

Let's leave aside complete immutability for a moment and consider a closely related issue that you can see in his code. He needs a map with defined iteration order, something I have needed occasionally in real world code.

He uses Go's builtin map type and a slice to define such a map. Now, how do you keep these two data structures in sync without restricting mutation? I think it's obvious that controlling mutation is a requirement here.

Go even admits that restricting access to data (i.e encapsulation) is necessary, and obviously preventing inconsistent mutation is a key reason why that is needed.

But then Go turns around and says, oh but you can't define fully featured data structures (i.e supporting range loops, type safety and indexing expressions) of your own that use encapsulation.

I have defended Go many times before, but the only way I can defend it is to say, yes it's true, this is a major weakness. We just don't know how to fix it without creating a lot of complexity elsewhere.