Gofumpt: A stricter gofmt | Hacker News Reader

Gofumpt: A stricter gofmt

github.com

98 pointsmetadat88 comments

Seems a lot of these rules focus on removing empty lines. I find empty lines let the code "breathe" a bit.

Without them, it's a massive wall of text, sometimes. Well, often times.

It's interesting you can address the "wall of text" problem in 2 ways.

1) Get IDE strategically insert vertical whitespace, for example right after function definition, and before function body. It does not need to be a full empty line, half of line height would do.

2) Or other way around, if you have an empty line before function body starts, IDE could narrow the height a bit, because this does not need a full-size empty line. Actually I wonder if there are IDEs around that play with variable line height for empty lines.

The above would allow folks who feel they are starved for vertical space to condense the code more, without impacting others who feel they have plenty of vertical space, and would like to break the "wall of text" with white space breathers.

Currently Intellij (GoLand) will strategically fold the `if / return err` pattern that Go code is littered with, so the whole "keep it on single line" point is moot for somebody using an advanced IDE that automatically renders the code in way to reduce visual clutter.

Empty lines let you split your code into assignment/initialization, logic and parameter validation, without requiring you to split everything into tiny functions.

I agree that empty lines can help. I've been careful to only remove empty lines which, in my opinion, don't help when structuring the code or making it more readable.

If you have examples where gofumpt is being a bit too aggressive, I'd like to hear about them. I've corrected, and even removed, some of gofumpt's rules in the past thanks to user input.

I use gofumpt for all my go code (have been for at least two years), and I use empty lines for structure. Out of the tens to hundreds of thousands lines written, I don’t recall a single instance where an intentional empty line was removed. I doubt gp’s comment is based on experience. Another giveaway is the “seems” at the very beginning.

Microsoft have an extension for Visual Studio that does this (and also shrinks lines with {}), I use it every day.

https://marketplace.visualstudio.com/items?itemName=VisualSt... (there’s been a version for the last few versions of Visual Studio too)

Put another way, writing clean code is like writing prose. You need to clearly break down your thoughtt, whether its sections, paragraphs, sentences, and fragments, or files, functions, blocks, groups of lines.

Drove me nuts when working in a C# code base that controlled newlines because I had to break up related thoughts.

The way you've described it is almost verbatim with how Google's C++ style guide puts it:

https://google.github.io/styleguide/cppguide.html#Vertical_W...

Yeah. And when I write comments, I do it in explanation of the chunk of lines below which perform a particular idea. The newline after that chunk is necessary to break up the chunks to give room for another one below, etc.

I really hate the c# code style of a new line for both open and close braces, especially in an if then else statement. Pushes people to use long ternary statements instead.

I prefer ifs to ternarys because, more often than not, I want to easily add and remove statements to the body of the then or else as I'm working things out.

Am I understanding correctly that: gofmt(gofumpt(code)) == gofumpt(code)?

If so, I don't get the negativity in these comments centered around diverging from gofmt. A developer could make any of the decisions that gofumpt made.

Yes, that's how it works.

If that's the case, then it's even more "gofmt" than "gofmt".

The important thing about "gofmt" is that it's the only canonical way to format an AST, so anything that fights with gofmt would undermine that and never get traction in the Go ecosystem.

Operating in the space where gofmt hasn't completely made up it's mind is a great way to experiment with potential improvements.

Good. Go deserves this.

Before gofmt, programmers were allowed to have accents. Teams were allowed to decide their preferred style.

Since gofmt, we "avoid bikeshedding" by disallowing the developer from adding any value via formatting.

Dozens of languages that are far more sophisticated than Go have adopted this boneheaded approach (typical Go philosophy) in the years since gofmt appeared.

It'll be a while before I retire, but when the time comes, I am preparing for this to be the hill I die on.

Ye. This is agile deep down removing the fun from programming with all its toxic processes. Programmers are the worst when it comes to forcing their dogmatic believes on others.

This, I really don't get why people want to enforce uniformity, a room filled with people with exactly the same accent is a boring room. Most teams I've worked on slowly converge to a "house style" in any case, and if you're on the team, your personal preferences are a part of that. Not some committee.

Version control. Auto-formatting means you'll never, ever, have to deal with diffs where 50% of the changes are indentation, commas and line breaks.

There is still room for 'style' in how you write code, what the code does is what matters. Formatting simply removes noise.

I agree with this 100%. I would extend it by saying that one you make formatting mandatory, the rules themselves don't matter, as long as they are the same for everyone on the project.

One could also make formatting automatic through git hooks and then every one can override their local formatting rules to their liking without breaking the common rules.

Eventually you realize that what's really needed is AST-based semantic revision control. Text is a lowest common denominator that all programming tools should go beyond, not just the compilers.

Nah, you still get those diffs when the autoformat rules change

Good thing Go and gofmt are very stable and consider compatibility a significant feature.

It depends on the formatter you're using. If it's strict about line width you'll end up with a lot of noise in the diffs. Prettier will re-format single-line lists that goes one character over the threshold and then re-condense them once it's able to fit under that threshold. It's not uncommon for a change that adds three classes to three different elements to become a 20 line diff.

Excellent.

Otherwise people sharing that criteria will never fix minor code style issues because "it was that way before, and it's such a minor thing, and it makes my merge harder, please accept the review and ship it don't be unreasonable", and of course we are "reasonable" and end up with thousands of "minor things" over time and the more there are the more people feel it's fine to leave shit in, the less comfortable they feel with the code until eventually they all quit and leave the mess to someone else.

I don't even want to imagine how bad the shitshow is going to get as AI coding assistants gain marketshare.

gofmt's normalization of code formatters was the single greatest achievement Go has to its name, and Prettier was a clear improvement on it. If fixing this crap is bothering anyone's laziness or their artistic sentimentalities then they better get to solving the issue where we're still storing code in source control instead of ASTs and letting the editors render them however the user wants.

I personally prefer a tool that formats because it avoid me have to go to meetings where I have to discuss coding style (which is even more boring).

"I really don't get why people want to enforce uniformity"

This is engineering, not food choice or movie accents. I don't want your second style in my code for the same reason that if my physical project uses metric tooling, I don't want the intern to come screaming in with imperial tooling, and then get huffy when I complain and ask them to recast their work in metric. Yes, it's true, there's absolutely nothing your imperial tooling can't do... but I still don't want it in my engineering for no benefit.

Engineering innovation is not harmed by being pushed above the basic level of which kind of tooling it uses.

I wonder how many people complaining are actually routine Go programmers. These rules are pretty sensible and clearly written by an experienced Go programmer. To the extent that my code would be modified by this tool, it isn't much. Many of the rules are just unsightly in Go code in practice and many of them I've already wondered why gofmt doesn't already have. For instance, the rule around newlines and brackets; I've often accidentally had a function end with a newline, but it never looks good, it's never something I'm like yes, let me fight for this and I've manually removed it as soon as I've seen it. If someone on my team really went to bat against these rules (and note the "really went to bat", I'm not talking quibbles here & there, we've all got those) I'd have questions about their ability to do basic cost/benefit analysis in such a process.

And, if it isn't engineering, then by all means go nuts with whatever you like. But I think your question only has meaning in a group context, because if you're not in a group context, who cares in the slightest what the non-existent group thinks? I've got plenty of personal projects that I in my own engineering role wouldn't consider to pass even basic engineering muster. You should see my electrical "engineering" work; I would feel dirty if I didn't scare quote the term "engineering". I'm not very good. Tends to look more like a hollywood bomb than anything you see linked in the sort of YouTube video that shows up on HN every so often.

Take PEP8 for example which is Python specific. PEP8 has a rule that code shall not exceed 79 characters. This has the added benefit that on a standard 1080 monitor you can see two Python files side by side without horizontal scrolling. This also has the added benefit of being able to see code diffs without issue.

I can see why certain choices are made but only matter if they have sound reasoning and arguments for them.

> This has the added benefit that on a standard 1080 monitor you can see two Python files side by side without horizontal scrolling.

Isn't that a function of font size rather than screen resolution?

You really want to argue that shrinking font from a standard font size will produce legible text like that?

The argument is against super long lines. I think Clean Code suggests something similar to the effect of if you have more than three arguments for a method you are better served taking a class as the argument.

you know that work is supposed to be boring, right?

This is actually about removing a number of chores from programming:

- having to worry about confirming to a style guide

- having to review and/or reformat other peoples code to conform to the style guide

- having to learn the style guide

- having to write a style guide

- having to break flow to format your code properly.

In what way is removing a bunch of gross chores intended to make code readable for others, is this about “toxic processes?”

gofmt and friends remove one of the most toxic processes of all: style guides.

I don't understand your comment. In the first sentences you seem to be in favor of code formatting (me too), pointing out the benefits, then towards the end you seem to be against it.

It was "go deserves this" where "this" is further punishment.

There's no accounting for taste to be sure, but I for one won't be patronized by a dumb tool. Formatting is for my team and me to decide, and no one else.

Write code as if the next maintainer is a psychotic murderer with your address.

Honestly I see gofmt as a lightweight pep8 (from python) and I prefer languages to follow a standard for style and such.

I kind of hate when I see people adopt a different style than the rest of the codebase moreso than anything else though. Keep things consistent, I dont want a zoo of inconsistent code styles.

Note that the first rule of pep8 tells us:

> A style guide is about consistency. Consistency with this style guide is important. Consistency within a project is more important. Consistency within one module or function is the most important. > However, know when to be inconsistent – sometimes style guide recommendations just aren’t applicable. When in doubt, use your best judgment. Look at other examples and decide what looks best.

Not everyone agrees with this, but just citing pep8 this is one of the more important points to me.

> disallowing the developer from adding any value via formatting

formatting doesn't add value. no one formatting is going to please everyone, so it follows that any formatting is going to piss someone off. So I think the point from the Go team is to remove that distraction and let people focus on whats actually important, the code itself.

Exactly. I prefer writing with double quotes as they are easier to hit for me. When/if the formatter changes them to single quotes I couldn't care less. I can accidentally miss-indent something and the formatter will catch it and fix it. Why spend mental energy making sure I'm perfectly writing trivial things when I can focus on the important stuff?

Someone never had to write a matrix

you mean a single dimension array pretending to be a multi dimensional array? no thanks.

What a lonely hill to die on.

Count me in to! There are dozens of us, dozens!

I personally am looking forward to the go tool replaces `go run` etc. for a version that makes additonal whitespace and misalignment into a compiler error!

this but actually earnestly yes

Not quite the same, but early versions of the gc compiler did run what is now gofmt on your code during compilation.

Anyone know if the following exists?

Gofmt/gofumpt, but as a special case, formats

    if err != nil { return err }

on a single line.

Not trolling, genuinely interested.

I've found that it's almost always a good idea to return fmt.Errorf(...), adding context to the error. Otherwise error messages are too generic and it's hard to figure out what's actually wrong. Then it doesn't fit on a single line anymore.

It doesn't? It is only a tiny tiny bit longer than the fmt.Errorf() by itself, which is surely one line, right? The key thing is to then have your syntax highlighter set up to specially handle these lines, to help you 1) know that they are there, 2) know that they are structured correctly, and 3) ignore them (which is safe now that you can see it exists and is in the right format). Unless you set these up as one line, your Go code gets littered with 3x (or even 4x if you add a blank line after each block, which starts to feel better) as much error handling lines as non-error lines.

I enjoy errors.Wrap/Wrapf too.

github.com/pkg/errors is dead — it’s been archived for a long time, and doesn’t incorporate go 1.20’s multi-error wrapping. Better switch to fmt.Errorf in new projects going forward.

fmt.Errorf with %w is how you do this since Go 1.13, gotta drop that dep

The default gofmt doesn't format this way, so any tool implementing this would not be compatible.

Does gofmt have a problem with error handling on one line?

You should really run gofmt on every file save.

"return err" is almost never the right thing to do.

Hate the padded block rule with a passion.

Maybe I'm weird but one of the greatest things about gofmt is that it doesn't murder "padded blocks", like so many formatters out there.

For example, given this

    func main() {

       fmt.Println("a")
       fmt.Println("b")


       fmt.Println("c")
       fmt.Println("d")

    }

gofmt will simply collapse the double line in the middle.

However, this tool will do this

    func main() {
      fmt.Println("a")
      fmt.Println("b")

      fmt.Println("c")
      fmt.Println("d")
    }

How is having chunks of code smashed into the tops and bottoms of a blocks good for improving readability? Why should the visual relationship between and main() and "a" be stronger than between "b" and "c" if the author doesn't want it?

As a non-Go developer, IMO the mistake is not putting the opening brace on its own line, for the “smashed into the top” reason you mention.

As far as I understand, that's a consequence of the automatic insertion of trailing semicolon by the lexer, not a decision by gofmt. It's explained here: https://go.dev/doc/effective_go#semicolons

I see. That doesn’t seem very sensible in potential-beginning-of-block contexts, in particular in the if case where an implicit empty statement is deduced after the condition.

> That doesn’t seem very sensible in potential-beginning-of-block contexts, in particular in the if case where an implicit empty statement is deduced after the condition.

Are you referring to C style if statements where you can elide the braces?

In go the braces are mandatory, so I don't think that really applies.

I was referring to the Go documentation linked above:

> One consequence of the semicolon insertion rules is that you cannot put the opening brace of a control structure (if, for, switch, or select) on the next line. If you do, a semicolon will be inserted before the brace, which could cause unwanted effects.

I was assuming that the “unwanted effects” are not simply a syntax error.

I hadn't really considered that people might want to do this. From experience reading and writing Go code for ~8 years at multiple companies, the only times I've seen leading or trailing empty lines in blocks have always been either inconsistent or unintentional, and usually both.

Are there Go codebases that stick to the formatting you show, out of curiosity?

Either way, please file a bug. Perhaps others can chime in there if they also use the same style.

I've seen this once across many Go codebases, it was someone who even a year in writing Go still refused to consistently name constants and global variables without snake_case, among other similar behaviors.

Support for allowing configuration of such a thing would be a misfeature if I was to be asked.

gofumpt will never have formatting knobs, following gofmt's design. But if one of gofumpt's rules forbids a style which is reasonable even if it's not very popular, we might want to make the rule more conservative or remove it entirely.

man you can't let snake_case pass code review

> How is having chunks of code smashed into the tops and bottoms of a blocks good for improving readability?

How is having a random newline at the top and bottom where there is already a clear separation between function signature and implementation via braces and the guide of the first indentation, helping readability?

All it does is steal valuable vertical space away from things that matter.

IMO, vertical space is a precious resource in software development. It shouldn't be squandered without clear and compelling reason.

I'm not saying that there's never a reason for to have extra blank lines. But your example isn't it.

there is no "visual relationship" between main and a, indentation differentiates blocks, empty lines at block begin/end serve no purpose, please stop doing this

Related prior discussions:

Gofmt No Longer Allows Spaces. Tabs Only - https://news.ycombinator.com/item?id=7914523 (2014, 56 comments)

Refactoring with go fmt - https://news.ycombinator.com/item?id=6724775 (2013, 40 comments)

Researchrsc: Gofmt - https://news.ycombinator.com/item?id=996934 (2009, 34 comments)

Tangentially related:

Psychological effects of coding style (2016) - https://news.ycombinator.com/item?id=22992914 (2020, 71 comments)