Open Discussion: What Tips Do You Have For Writing Clean Code?

I read that article out of curiosity and I have to say, I am more than just a bit irritated by it.

The article is a perfect example of someone selling a personal preference as if its a fact which everyone should obey.

The autor even admits that after doing some research he found out that just about 10% of other developers follow his example and the majority does logical grouping.

I have to say, I think alphabetical oordering is just as bad as random ordering. The two examples looked exactly the same for me. I am very much in favor of the logical grouping (with blank lines in between!) and think its much easier to scan over.

But still: is is preference and not "clean code".

If you think about improving your general coding abilities, I can warmly recommend two books:

A philosophy of software design
by John Ousterhout

Refactoring (second edition)
by Martin Fowler

I learned a lot out of these two books.

I like to think of this in radiating bits of advice. First rule: no methods may be longer than 10 lines. From that rule many things follow.

I then look to Sandi Metz's guidance:

1. No class must be greater than 100 lines
2. No method must be greater than 5 lines
3. No more than 4 parameters in a method
4. Controllers can instantiate only one instance variable

Or look to literate programming. Or command/query separation. Or the Law of Demeter.

But focus on small methods and these things begin to emerge.

1. Don't be lazy about typing. Abbreviations harm readability.
2. Don't encode data in your control flow. switch/case is evil.
3. Don't implement what you need right now. Think in abstracts.

I recently wrote all of these rules down for myself. The document is more than 600 lines long, so here are my 3 most important tldrs. All of them are inspired by Uncle Bob's best seller.

1. Write code so it resembles its own specification

I find this the most relevant point, but also the hardest. I got the idea by combining the rule 'Use intention-revealing names' from Clean Code and 'Start Statements with Verbs' and 'Be brief' from Microsoft's writing guidelines.

A bit easier to imagine is: Write code like you write a newspaper article. A common rule here is The Pyramid Principle of writing. So all those essays in highschool were actually good for something

2. Prefer readability over performance

You might have noticed there's a pattern already. Code must be read by people, not machines. In most of the cases, the fancy JS 1-liners come with a trade-off in readability. MS writing guidelines support this - 'Write like you speak' tells you to write like a person, not a machine.

3. Do proper error handling and always do it the same way

The idea for this comes from the chapter 'Use special error objects'.

My solution is to use try-catch inside a custom function that always returns an error-tuple.

In Typescript, the type looks like this:

type TryCatchTuple<T> = [T | null, Error | null];
// Resolves in a 2-element array with [result, error]
// -> result is null if there's an error
// -> error is null if there's a result

It's used in this function (async example):

const tryCatchAsync = async <T>(fn: () => Promise<T>, ): Promise<TryCatchTuple<T>> => {
    try {
        const result = await fn()
        return [result, null];
    } catch (error) {
        return [null, error];
    }
}

For sync, just leave out the async/await and the Promise generics.

1. Remember "Clean Code" is subjective
2. Don't prioritise "cleanliness" over efficiency

To me, it's usually a triangle: Efficiency, readability, abstraction.
However, it's rarely the same distribution, as it can vary from one project to another, depending on the context. I can't give a general rule but this how I see the project in terms of clean code.

You can't neglect performances but you have to avoid micro-optimization, especially when the code becomes too "ninja", a.k.a. less readable and maintainable. If you're the only one who can modify the script, it can become a problem quickly.

Too much abstraction is not really good in my experience. It can jam code reviews significantly. Besides, you can get a lot of unwanted side effects, so more stress for maintenance, etc.

About the article, I feel that is more personal taste than some pragmatic thing. There is a big difference between opinion (which everyone has) and well-known metrics such as SLOC, Cyclomatic Complexity and so on.

As tips, I suggest reading those books:

• Growing Object-Oriented Software, Guided by Tests
• The Pragmatic Programmer
• 99 Bottles of OOP
• Refactoring
• Domain Driven Design
• Extreme Programming Explained

As a general rules of thumb that I like to follow:

• Follow Tell, Don't Ask when possible, but not become too dogmatic about that
• Law of Demeter is not a law, but it helps drive the code to a good and maintainable future
• Start writing acceptance (e2e) tests, then write unit/integration tests, then write the code. This helps a lot organizing ideas and communication between the parts, without writing all the base classes upfront.
• ReadMe driven development help organize what to do first and this impact directly to the code organization
• Know your tools, this help you to use your tools to write your code.
• Use dependency injection, even if it needs to be done in a manual way.
• Write code for humans, not for machine.
• (questionable taste) Switch/case isn't evil. Use it in factories to help to use polymorphism
• (questionable taste) Do not put repeated prefixes or suffixes, like "UseCase", "I", "Service", "Entity" and so on (unless if needed to work, like the prefix use in React Hooks). Each object should tell a story, each object should deserve its place in the source code by the story that it tells, not by its name.

Yes, definitely. The rule of thumb needs to be: quick to understand. If too complicated, break it up.

Imagine while coding the colleague who has to maintain the code after you is a violent psychopath who knows where you live. 😅

To me, a clean code rule has to be something that is worth the effort in the end.
I'm not that fast at alphabetical sorting, so I would take a lot of extra time and it would exceed the benefits it could provide.
My favourite type of sorting is the one my IDE does on autoformat.

The more articles I read on "clean code", the more it reminds me of a religion or a cult :)

hmm.. I don't like that approach. I've never tried it but I feel like that would slow me down. I preffer modular and small css files. Another thing I like doing is using the id attribute to distinguish the different sections. So I can write styles like so #info_card .title which makes it easier to find the section I need to work in.

Writing CSS properties in alphabetical order? I fail to see how this can make the code clearer; maybe it helps when you are searching for a property, but Ctrl-F (or Ctrl-s if you are an emacsian) works fine too.

To me "clean code" means code that makes it easier to understand what its goal is and how it achieves it. In this respect, I find that using meaningful names for variables, functions, etc. helps a lot. Using some comment to explain the goal of the function, module, etc, and the "conceptual model" behind it helps a lot too (and this is a kind of comment that is unlikely to get stale, since it is related to the overall architecture). Comments explain the algorithm can be helpful if the algorithm is not trivial and/or to explain some choices that seem strange at first sight (e.g., "We use this structure instead of the more obvious ... because later we need to ...")