DEV Community

Lane Wagner for Boot.dev

Posted on • Originally published at qvault.io on

The Pros and Cons of DRY Code

desert landscape

The post The Pros and Cons of DRY Code first appeared on Qvault.

Clean code is like clean garbage – it’s only truly clean if it doesn’t exist. In other words, the only clean code is no code. Let’s start with an acknowledgment that a perfectly clean (empty) codebase is useless. With that in mind, our pursuit of clean code will necessarily consist of tradeoffs. We’ll trade usefulness for cleanliness, complexity for speed, ownership for ease of development, and abstractions for reusability. DRY (don’t repeat yourself) code is often held aloft as an ideal in the quest for clean code. Let’s explore why I think DRY can be a good heuristic, but far from an absolute.

What is DRY Code?

According to Wikipedia:

Don’t repeat yourself is a principle aimed at reducing repetition of software patterns, replacing it with abstractions or using data normalization to avoid redundancy. Every piece of knowledge must have a single, unambiguous, authoritative representation within a system.

While this definition is fairly exhaustive, what I’ve found people usually mean when they say “DRY code” is that when you find yourself repeating pieces of logic you should instead create a reusable function, interface, class, etc, so that you only have to maintain one copy of it. For example, take the following API calls:

export async function updateUserHandle(handle) {
  if (!isLoggedIn()){
    // redirect to login screen
    return;
  }
  let token = localStorage.getItem(jwtKey);
  let decodedToken = decodeJWT(token);
  const hoursDelta = 24;
  if (decodedToken.exp < (Date.now() + hoursDelta*60*60) / 1000){
    refreshToken();
  }
  return await fetch(`${domain}/v1/users/handle`, {
    method: 'PUT',
    mode: 'cors',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${token}`
    },
    body: JSON.stringify({
      handle
    })
  });
}

export async function updateUserInterests(interestUUIDs) {
  if (!isLoggedIn()){
    // redirect to login screen
    return;
  }
  let token = localStorage.getItem(jwtKey);
  let decodedToken = decodeJWT(token);
  const hoursDelta = 24;
  if (decodedToken.exp < (Date.now() + hoursDelta*60*60) / 1000){
    refreshToken();
  }
  return await fetch(`${domain}/v1/users/interests`, {
    method: 'PUT',
    mode: 'cors',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${token}`
    },
    body: JSON.stringify({
      interestUUIDs
    })
  });
}
Enter fullscreen mode Exit fullscreen mode

You may have noticed, but the beginning of those two API calls are nearly identical – they handle some basic logic that checks to see if the user is properly authenticated and sends that authentication in the request. This might not be a big deal with just two API calls, but what if we have 30? Or maybe 1000? Instead, we can DRY up this code by writing a simple fetchWithAuth function that will centralize all the client’s authentication logic in a single place:

async function fetchWithAuth(url, params){
  if (!isLoggedIn()){
    // redirect to login screen
    return;
  }
  let token = localStorage.getItem(jwtKey);
  let decodedToken = decodeJWT(token);
  const hoursDelta = 24;
  if (decodedToken.exp < (Date.now() + hoursDelta*60*60) / 1000){
    refreshToken();
  }
  if (!params.headers){
    params.headers = {};
  }
  params.headers.Authorization = `Bearer ${token}`;
  return await fetch(url, params);
}

export async function updateUserHandle(handle) {
  return await fetchWithAuth(`${domain}/v1/users/handle`, {
    method: 'PUT',
    mode: 'cors',
    headers: {
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      handle
    })
  });
}

export async function updateUserInterests(interestUUIDs) {
  return await fetchWithAuth(`${domain}/v1/users/interests`, {
    method: 'PUT',
    mode: 'cors',
    headers: {
      'Content-Type': 'application/json'
    },
    body: JSON.stringify({
      interestUUIDs
    })
  });
}
Enter fullscreen mode Exit fullscreen mode

Why wouldn’t you DRY out your code?

Seems like a good idea to reduce code duplication right? Well, yes, in general it is. That said, lets look at some of the drawbacks that come along with too much centralization.

1. Too many abstractions

Sometimes code is the same at a given point in time, but later on, it becomes distinct in some way. It’s really hard to guarantee that duplicate code will remain perfect copies of each other forever. A hypothetical example of this would be if the Facebook and Instagram APIs had the exact same way to create a social post. Just because they’re coincidentally the same, probably doesn’t mean that the logic should only be written once.

The solution is likely to remain disciplined about splitting out code that, while it may be similar, is only coincidentally similar. Try to only merge code that’s fundamentally the same. A great example would be a math function like log2. That function should work for every case that you need to calculate a logarithm – each call is fundamentally the same.

2. External dependency creation

If two different projects share the same logic, it can often make sense to centralize it in a library package. While this is often a great idea, it can add overhead and can end up being more trouble than it’s worth. For example, even if the abstraction makes sense, you’re definitely adding at least the following complexity to the project:

  • Management of the dependencies versions and running updates regularly
  • Requires multi-project updates in order to get a new change to a specific dependent
  • Often involves more remote infrastructure like NPM or PyPy
  • Gets harder to make “breaking” changes to the libraries core functions – requires a higher standard of code quality and architecture

3. Localization complexity

When debugging or reading code, it would be easiest if the flow of logic started at the top of a file and flowed in a linear path. For example:

START PROGRAM
INSTRUCTION 0
INSTRUCTION 1
INSTRUCTION 2
INSTRUCTION 3
END PROGRAM
Enter fullscreen mode Exit fullscreen mode

Unfortunately, in useful programs, we need functions, classes, methods, type definitions, etc. that force us to read and write code in a non-linear way. The goal should be to keep everything as linear as possible and sacrifice linearity and simplicity for reusability and separation of concerns as necessary. Every time we extract a chunk of code from a larger function into a smaller more encapsulated one, the code becomes just a little bit harder to follow.

With a highly compartmentalized project, when we see a function called getUser, if we want to really know what’s going on we have to peek into that function and remember the current context because we’re now looking at a different file. Obviously, the cognitive burden becomes greater and greater the more definitions you need to jump through.

Takeaways – Code smells and hueristics

Since no code is perfect, we can need to make use of some heuristics (rules of thumb), to try to work towards a cleaner codebase.

1. WET code, or the rule of three

WET is a better rule of thumb in my opinion than DRY.

WET stands for “write everything twice”, and forces you to think a bit harder about whether or not a piece of logic deserves an abstraction. The rule of three is an alternative that says you should even wait until you’ve written something three times before breaking it out.

2. Is it testable?

Most functions should be predictable and testable. They should behave like math functions or pure functions) – given a set of inputs you’ll always receive the same outputs, and the state of the program isn’t mutated. If the code you’re thinking about condensing into a function can be a pure function, then it’s likely a better idea than if it were an impure function.

Pure functions are really easy to write good unit tests for – if your abstraction is easily testable it’s more likely to be a good abstraction.

3. Are there special cases or arguments only used by a fraction of callers?

Take a look at the following example:

function getArea(height, width){
  return height * width
}
Enter fullscreen mode Exit fullscreen mode

This is a great function! Its very simple, and obviously can be used to calculate the area of any shape. Here’s a bad example:

function getArea(height, width, isTriangle){
  if (isTriangle){
    return (height * width) / 2
  }
  return height * width
}
Enter fullscreen mode Exit fullscreen mode

Special cases are bad news – I’m trying to be too abstract. Instead I should just create two separate functions:

function getTriangleArea(height, width){
  return (height * width) / 2
}

function getRectangleArea(height, width){
  return height * width
}
Enter fullscreen mode Exit fullscreen mode

Have any more ideas for good rules of thumb? Let me know on Twitter! I’d love to update and improve this article.

Thanks For Reading!

Take computer science courses on our new platform

Follow and hit us up on Twitter @q_vault if you have any questions or comments

Subscribe to our Newsletter for more programming articles

Top comments (2)

Collapse
 
pinotattari profile image
Riccardo Bernardini

I do not have a "decision algorithm" for deciding if "factorizing out" a piece of code or not, but roughly I would say that I apply DRY if at least one of the following conditions is satisfied

  1. The action has a meaning by itself. For example, the code necessary to download a file after authentication could be implemented as an independent piece of code.
  2. The duplicated code is quite complex and long and/or the probability of updating it in the future are quite large.

Actually, sometimes I do kind of opposite and put in code procedures that are maybe just called once in to improve readability. Sometimes it can happen that a procedure is quite long and does lots of different actions; in order to make some "order" I split the code in different procedures (with nice descriptive names). In my opinion this makes the code easier to understand. This is quite convenient in Ada (the language I code with) since you can define procedures inside other procedures, limiting the visibility (and the need for hunting) of the "readability procedures."

Collapse
 
ronnewcomb profile image
Ron Newcomb

I also sometimes do/don't factor out a piece of code based on whether I have a good, obvious name for it.