That was probably ~80% of why I opted out as well (the rest ~20% are mutability and bad parallelism story).
> Feels too much like macros and C++ templates. I just don't want to be that smart.
I don't even view it as being smart. It's more like a specialised area of expertise that you need to grind through in order to be productive, and even when you're an expert it's still error-prone. Turns out though, a lot of other tech doesn't require that so, like you, I moved there.
> I keep going back to Bill Joy's goal of enabling working both in the big and in the small at the same time.
Admirable goal. I try to do the same but it's not an easy thing to achieve. What language would you recommend in this area?
> I could focus more on design and less on wrestling the compiler.
Depends what you mean. I am very okay to fight with Rust's compiler because that way I know I won't have dangling pointers or references and will not inadvertently try to modify the same piece of data from multiple threads. I am also very okay with fighting with OCaml's compiler because I often found myself not thinking properly about the types of my code.
But in C and C++, yeah, there the fight is kind of just "please compile and leave me alone" and it's not that much of "it's compiled, therefore you get guarantees X and Y".
> But I wish language (and stack) designers also removed features. Cull deprecated stuff. More generalizations. Fewer primitives. Fewer classes. Etc.
Absolutely! I wish more language designers/maintainers had the courage to do this. They also don't have to do it in a breaking way; they can just extract the deprecated (and removed) functionality into an optional dependency that people can use if they are so hell-bent on not using the officially sanctioned way from version X.Y and on.
But I guess even that would catch many people by surprise and provoke negative reactions so it's probably a very strong reason on why it isn't being done.
So what's left? A linter that's invoked like this: `idiomatic-rust --check --strict`?
I guess that's the only way out for now.
I wish I knew.
For ages, I've imagined a Java do-over I code named Encore. What would "Java" look like today, given the same original goals of improving programmer productivity and reducing errors.
But now I've switched tacts almost completely.
TLDR: Code generation 2.0. Code you'd write yourself, if only you had more time.
So with SQL, for example, instead of a wrapper or model which generates SQL at runtime, start with the SQL and generate code (in your target language) at compile time.
I've done this, with varying degrees, for SQL, HTML, HTTP, HL7, DICOM. Just trying to flesh out the idea, strategy.
One consequence is the target language becomes much less important. So now I feel (hunch, as in "believe but cannot yet prove") that the smaller the language, the better. Like BASIC, Pascal, Lua (which I don't particularly like) small. Barely more than a scripting language with some objectness and modules.
And even more impactful, the contemporary libraries and frameworks become much less important, or even moot.
Thanks for replying. I'll post something if my efforts mature enough for sharing.
Oh Hell yes. I wish they have plan for depreciation. Trial, > Warning > Error > Removed. Push the whole language and ecosystem forward. That is 10 times easier then going to a new languages and making new ecosystem.
This is why I opted out of Ruby. Feels too much like macros and C++ templates. I just don't want to be that smart.
I keep going back to Bill Joy's goal of enabling working both in the big and in the small at the same time.
Mid 90s, Java was logical step from C-ish towards Scheme-ish and Simula-ish (SmallTalk, Self). I could focus more on design and less on wrestling the compiler. Much progress since then, in Java and elsewhere.
But I wish language (and stack) designers also removed features. Cull deprecated stuff. More generalizations. Fewer primitives. Fewer classes. Etc.