Sometimes I think Go was specifically made for Google to dictate its own preferences on the rest of us like some kind of power play. It enforces one single style of programming too much.
Ew, that’s awful. Go is not one of my programming languages but I had always held it in high esteem because Ken Thompson and Rob Pike were involved in it.
Honestly, it does not happen often that I have a ln unused variable that I want to keep. In my mind it is the same thing when wanting to call a function that does not exists. Also my editor is highlighting error Long before I try to compile, so this is fine too for me.
Worse than not having a unused variable check at all? Dunno, the underscore assignment are very visible for me and stand out on every code read and review.
Yes, worse, because now if you want to use the underscore assignment to indicate that you really want to discard that variable - it gets confused with underscore assignments that were put there “temporarily” for experimentation purpose.
Say I’m having some issue with a function. I comment out half the function to see if that’s where the weirdness is. Golang says “unused variable, I refuse to compile this dogshit!” I completely fool Golang by just using _ = foo. Yes, I was correct, that’s where the problem was. I rewrite that section of the code, and test it out, things work perfectly. Only now, it turns out I’m not using foo anymore, and Golang has no idea because I so cleverly fooled it with _ = foo.
Now, something that could be caught by a linter and expressed as a warning is missed by the language police entirely, and may make it into production code.
Police the code that people put into a repository / share with others. Don’t police the code that people just want to test on their own.
The underscore is used in production code too. It’s a legitimate way to tell the compiler to discard the object because you don’t intend to use the pointer/value.
From what I’ve heard from Google employees Google is really stringent with their coding standards and they usually limit what you can do with the language. Like for C++ they don’t even use half the fancy features C++ offers you because it’s hard to reason about them.
I guess that policy makes sense but I feel like it takes out all the fun out of the job.
Just about any place I know that uses C++ also does that with C++ so that’s nothing unusual for C++ specifically. It’s too big of a language to reason about very well if you don’t, so you’ve gotta find a subset that works.
Too many patterns. If you do not do this every author will have a different use of the language and you will have to read a book of documentation each time you change files.
I think this is a good thing. The styles are just opinions anyway and forcing everyone to just follow a single style takes a lot of bikeshedding away, which I really like.
The best part of these threads is no matter what someone comments, at least 2 people will reply either correcting or “clarifying” the original commenter.
A quick “find all references” will point out it’s not used and can be deleted if it accidentally gets checked in but ideally, you have systems in place to not let it get checked into the main branch in the first place.
Or a linter. Or code reviews. Or anything else. The nice thing is that if the compiler doesn’t demand something, it can be given to the engineer as an option. The compiler should have the option to do it. The option could even be defaulted on. Afaik there is no way in Golang to disable that error (this is the line that does it: github.com/golang/go/blob/…/stmt.go#L67-L69). like --no-pedantics or such. Golang’s compiler openly refuses to give engineers more choices in what they think is the best system to handle it.
If it is a pure value, I’d assume yes, but if it is tied to a side effect (E.g. write its value to a file), then it would be not used but still could break your app if removed.
I’m not familiar with rust language specifically, but generally that’s what could happen
I prefer for it to be just a warning so I can debug without trouble, the build system will just prevent me from completing the pull request with it (and any other warning).
If only there was some way the compiler could detect unused variable declarations, and may be emit some sort of “warning”, which would be sort of like an “error”, but wouldn’t cause the build to fail, and could be treated as an error in CI pipelines
Let’s not pretend people acknowledge warnings, though. It’s a popular meme that projects will have hundreds of warnings and that devs will ignore them all.
There’s a perfectly valid use case for opinionated languages that don’t let you get away with that. It’s also similar to how go has gofmt to enforce a consistent formatting.
You can, if you want, opt into warnings causing your build to fail. This is commonly done in larger projects. If your merge request builds with warnings, it does not get merged.
In other words, it’s not a bad idea to want to flag unused variables and prevent them from ending up in source control. It’s a bad idea for the compiler to also pretend it’s a linter, and for this behaviour to be forced on, which ironically breaks the Unix philosophy principle of doing one thing and doing it well.
Mind you, this is an extremely minor pain point, but frankly this is like most Go design choices wherein the idea isn’t bad, but there exists a much better way to solve the problem.
Please ignore everyone else being unkind - I’m somewhat new to build systems in general, what are the advantages/disadvantages of Bazel compared to other build systems?
It's very funny for sure and I was just about to share it with the other SWEs in our off-topic chat. But I just can't listen to it, had to close the tab after like 50 seconds. Perhaps on mute and with CC?
programmer_humor
Active
This magazine is from a federated server and may be incomplete. Browse more on the original instance.