Collection of code simplification and refactoring tips #best practices #refactoring #code review

README.md

Collection of code simplification and refactoring tips

Want to write clean code? Well here's your chance. The following set of examples are practices that I've come across during my times as a programmer and that I believe in. At least at the moment. Please feel free to add comments, criticise my thinking and contribute to everyone's knowledge. I'm trying to find a better platform for these, but for now a gist should do. I'm doing this solely to document, analyse and share my own thinking, which oftentimes is quite unstructured. These days I'm not too fussed about the syntactical structures of the code, but would rather concentrate on painting the largest guidelines and deciding the direction I want my programming style to move towards.

Contents:

• Define value boundaries early, keep things flat
• On useless try-catches, being overly defensive, I/O boundaries and variable scope
• Know what you're returning and be explicit about it
• Avoid unnecessary ternaries
• ...

Further reading:

• 23 guidelines for writing readable code
• Clean Code concepts adapted for JavaScript

IDEA: backend: no http stuff to service layer, no database stuff to http layer (generalised)

Like "function A calculates api path", "function B does the fetch" rather than everything in a bag.

IDEA:

• no configuration objects to functions
• max 3 parameters per function
• calculatePos(x1, y1, x2, y2) can be simplified by creating a Line type or a Point type - telegram convo 4.10

Shorter if branch first

This works in most cases, however I wouldn't only leave it as the end result is, but would combine it with Define value boundaries early, keep things flat.

If statements blocks with more than few lines of code in them are usually a bad sign!

From:

if (something) {
  /*
   *
   *
   * Loads of stuff here so you need to scroll
   *
   *
   * ...
   */
} else {
  // Only one or few lines
}

To:

if (!something) {
  // Only one or few lines
} else {
  /*
   *
   *
   * Loads of stuff here so you need to scroll
   *
   *
   * ...
   */
}

Always avoid unnecessary data structures, stop creating garbage piles for your data!

Consider the following JSON response from an HTTP API:

// GET users/123
{
  data: {
    username: 'rikukissa',
    level: 'noob',
    info: {
      age: 26,
      email: '...'
    }
  }
}

There's at least 2 different things wrong here: First of all the data wrapper is pretty much useless. Of course you could argue that there's need for meta data, error information or what ever, and that's a perfectly valid claim. In this case though, there's no information of that type included in my response. One could also argue, that error information should only be a part of erroneous responses and not valid ones so something like:

// GET users/123
{
  error: false,
  data: {
    username: 'rikukissa',
    level: 'noob',
    info: {
      age: 26,
      email: '...'
    }
  }
}

is completely pointless.

Stick to convention, always

There are even wilder examples like { data: [], errors: [], meta: [] }, and the question is: what kind of information would you store in the meta field? I bet you a pint all suggestions you'll come up with would be better off transferred in some other part of the HTTP protocol like in the response headers. My worry is that these unconventional responses are in reality only used because of the lack of knowledge about best practices and the features HTTP provides by default. There are for instance very few error cases that couldn't be mediated with only HTTP response code.

Don't create places to dump code

Okay, let's move on. This is what we have now:

// GET users/123
{
  username: 'rikukissa',
  level: 'noob',
  info: {
    age: 26,
    email: '...'
  }
}

It looks fairly good, although the need for a separate info section puzzles me. There are many names for these kinds of sections like meta, info, data, recycle bin, garbage, utils, common. Let's face it: it's just a unstructured place where you can dump your new field without re-thinking the overall structure. If we were to model the response as a data type User, we'd most likely not treat "age" any different from the rest of the fields

interface IUser {
  username: string;
  level: string;
  age: number;
  email: string;
}

You may have noticed that the name examples I gave such as utils and common aren't really related to this specific use case. And there's a reason for this: people tend to do this everywhere. A directory in your codebase called common is just a telltale sign of bad design. There are very few modules that are "common" for all other modules of your codebase. I feel like I always say this, but to me it's the same as creating a file called functions.js with all the functions of your app inside of it. Please don't.

IDEA: Prefer passing data instead of functions as parameters

• Special cases: high-level abstractions like map etc functions, callbacks..
• Logic upwards propagation argument

Example

From this:

getDynamicSelectOptions = (field: IField, values: IFormData) => {
  if (field.dynamicOptions) {
    switch (field.name) {
      case "city":
        return addresses[values.state].cities;
      case "cityPermanent":
        return addresses[values.statePermanent].cities;
      case "addressLine4":
        if (
          addresses[values.state][values.city] &&
          addresses[values.state][values.city].areas
        ) {
          return addresses[values.state][values.city].areas;
        } else {
          return [];
        }
      default:
        return [];
    }
  } else {
    return [];
  }
};

To this:

getDynamicSelectOptions = (field: IField, values: IFormData) => {
  if (!field.dynamicOptions) {
    return [];
  }
  switch (field.name) {
    case "city":
      return addresses[values.state].cities;

    case "cityPermanent":
      return addresses[values.statePermanent].cities;

    case "addressLine4":
      const hasAreas =
        addresses[values.state][values.city] &&
        addresses[values.state][values.city].areas;

      if (!hasAreas) {
        return [];
      }

      return addresses[values.state][values.city].areas;

    default:
      return [];
  }
};

Combination of:
Define value boundaries early, keep things flat
Shorter if branch first

The code's still quite difficult to read, but more about that later

Thanks for the useluf tips. I especially like the 'Define value boundaries early, keep things flat' tip! I'll try to remember that.

Oh, thanks a lot for your feedback and for taking time to read through some of it :)

How many levels of module dependency hierarchy is okay?

👍 Avoid creating dependency hierarchies deeper than 4 levels. Aim for 2-3.

Few times during my career I've come across projects with +10 levels of dependant modules. Most of them have been Java projects.
The issue is, code like that is extremely tiring to follow. I remember writing down the path I had taken whilst reading the code just so I won't forget how I ended up to the current module I'm reading. Not good.

I feel like the best way to approach this would be to try and describe the whole process from start to finish on the main level:

function getCake() {
  oven = getPreheatOven()
  pan = getGreasedPan()
  mixture = mixIngredients(getEggs(), getFlour())
  putMixtureIntoPan(pan, mixture)
  return bakeForMinutes(40, oven, pan)
}

As the alternative (bad) approach I see functions that instead of being written to flow from top to bottom,
are written to flow deeper and deeper the module tree

function getCake() {
  return getBakedCake(40)
}

function getBakedCake(minutes) {
  oven = getPreheatOven()
  pan = getPanWithMixture()
  return bakeForMinutes(minutes, oven, pan)
}

function getPanWithMixture() {
  pan = getGreasedPan()
  mixture = mixIngredients(getEggs(), getFlour())
  return putMixtureIntoPan(pan, mixture)
}

... and so on. The reader would have to go through 3 different files just to know what ingredients are used.

How to solve cyclic dependencies between modules

AKA "why am I getting undefined when I import something from my module

Hey! A classic situation with a very generic and bad example (TODO). You've written few modules and soon notice that they depend on each other for one reason or another. Albeit being a **very strong indicator of bad design**, this is a common situation and one that's easy to solve:

user.ts

import { calculatePoints } from './points'
  
interface User {
   createdAt: number; // in milliseconds
   level: number;
}

export function updateUserPoints(user) {
  return {
    ...user,
    points: calculatePoints(user) // dependency to points.ts
  };
}

export function getUserRegisteredInYears(user) {
  return Math.floor((Date.now() - user.age) / 1000 / 3600 / 24 / 365); // Please, don't to this in real life
}

points.ts

import getUserAge from './user'
  
export function calculatePoints(user) {
  return getUserRegisteredInYears(user) * 3; // dependency to user.ts
}

There are few reasons why this happened in the first place:

• Responsibilities of each module are too loosely defined
• There are more concepts in play that you first imagined

And few solutions:

Responsibilities of each module are too loosely defined

Is there really a need for a point module? If not, the user module could look like this:

interface User {
   createdAt: number; // in milliseconds
   level: number;
}

function calculatePoints(user) {
  return getUserRegisteredInYears(user) * 3;
}

export function updateUserPoints(user) {
  return {
    ...user,
    points: calculatePoints(user)
  };
}

export function getUserRegisteredInYears(user) {
  return getYears(Date.now() - user.age);
}

However, if you answered "yes", could it be that the points module knows a bit too much about the internals of the user module? Could we, for example, turn the signature of calculatePoints into being:

- export function calculatePoints(user) {
+ export function calculatePoints(userAgeInYears) {

which would simplify the function to:

export function calculatePoints(userAgeInYears) {
  return userAgeInYears * 3;
}

There are more concepts in play from what you first imagined

In this very simple example, the only thing that the getUserRegisteredInYears function really provides us with is the conversion from milliseconds since the creation of the user to years. Could we add another module called time.ts? That way we could drop the getUserRegisteredInYears function altogether and refer to this new module from our points module:

points.ts

import getYears from './time'
  
export function calculatePoints(user) {
  return getYears(Date.now() - user.createdAt) * 3;
}

Stop repeating yourself!

👎

interface User {
 userCreatedAt: number;
 userPoints: number;
}

👍

interface User {
 createdAt: number;
 points: number;
}

Repeating the word user creates extra fuzz around the words that are actually relevant, without adding any value. The reader should be able to infer the context even when it's few lines above.

---

👎

it('should add two numbers together', ...)
it('should fail if divider is zero', ...)

👍

it('adds two numbers together', ...)
it('fails if divider is zero', ...)

By rephrasing the test case description, we can keep it as descriptive while making it more concise. Someone could even argue, that a test not only verifies that your code should do something, but that it actually does the thing it was written for.

Don't add unneeded context

Always name your constants

Got this idea while writing the bit about cyclic dependencies. You might have come across time calculation like this before:

// Returns true if token is older than 3 days
export function hasTokenExpired(token) {
  return (Date.now - token.createdAt) > 86400000 * 3;
}

The comment is pretty spot on, but we can do better:

export function hasTokenExpired(token) {
  const day = 24 * 3600 * 1000;
  return (Date.now - token.createdAt) > day * 3;
}

Not sure if you disagree, but to me it's much easier to read that something's greater than day * 3 (probably means 3 days) than that something's greater than 86400000 (possibly the mass of the Earth)

Great quote from clean-code-javascript:

We will read more code than we will ever write. It's important that the code we do write is readable and searchable. By not naming variables that end up being meaningful for understanding our program, we hurt our readers. Make your names searchable.