Comment by the_gipsy
21 hours ago
> This is why unused dependencies are a compile time error.
https://go.dev/doc/faq?utm_source=chatgpt.com#unused_variabl...
> There are two reasons for having no warnings. First, if it’s worth complaining about, it’s worth fixing in the code. (Conversely, if it’s not worth fixing, it’s not worth mentioning.) Second, having the compiler generate warnings encourages the implementation to warn about weak cases that can make compilation noisy, masking real errors that should be fixed.
I believe this was a mistake (one that sadly Zig also follows). In practice there are too many things that wouldn't make sense being compiler errors, so you need to run a linter. When you need to comment out or remove some code temporarily, it won't even build, and then you have to remove a chain of unused vars/imports until it let's you, it's just annoying.
Meanwhile, unlinted go programs are full of little bugs, e.g. unchecked errors or bugs in err-var misuse. If there only were warnings...
Yeah, but just going back to warnings would be a regression.
I believe the correct approach is to offer two build modes: release and debug.
Debug compiles super fast and allows unused variables etc, but the resulting binary runs super slowly, maybe with extra safety checks too, like the race detector.
Release is the default, is strict and runs fast.
That way you can mess about in development all you want, but need to clean up before releasing. It would also take the pressure off having release builds compile fast, allowing for more optimisation passes.
That doesn't make any sense, you'd still need to run the linters on release. Why bail out on "unused var" and not on actually harmful stuff.
> Debug compiles super fast and allows unused variables etc, but the resulting binary runs super slowly, maybe with extra safety checks too, like the race detector.
At least in the golang / unused-vars at Google case, allowing unused vars is explicitly one of the things that makes compilation slower.
In that case it's not "faster compilation as in less optimization". It's "faster compilation as in don't have to chase down and potentially compile more parts of a 5,000,000,000 line codebase because an unused var isn't bringing in a dependency that gets immediately dropped on the floor".
So it's kinda an orthogonal concern.
Accidentally pulling in a unused dependency during development is, if not a purely hypothetical scenario, at least an extreme edge case. During debug, most of the times you already built those 5000000000 lines while trying to reproduce a problem on the original version of the code. Since that didn’t help, you now want to try commenting out one function call. Beep! Unused var.
Right, I meant that the binary should run slowly on purpose, so that people don't end up defaulting to just using the debug build. A nice way of doing so without just putting `sleep()`s everywhere would be to enable extra safety checks.