A well-architected web application views its UI as a mere layer or "skin" over the core business logic.
UI are a way for users to interact with the underlying logic, but it doesn't need
to influence the logic itself.
Let's take this a step further: could a user perform all operations in your web application from a console, if you decided to build one?
This can serve as a good litmus test in my opinion.
If you cant let users do the same actions they do in your web application from a console and vice versa,
you are not separating your logic enough.
Why it's important?
There are several advantages of separating UI and business logic.
Maintenance and Debugging Bugs and user flows can be isolated and fixed without going through complex steps.
Testing You can test the UI and the logic independently, making tests more accurate and
easier to write.Reuse code Isolated pieces can be reused in different parts of your application or even across applications, saving coding time and preventing duplication.
Replaceability You can update separate pieces without rewriting others. this leads to your application being adaptable to changing requirements, which is important in our agile world.
Let's see how those advantages are achieved in practice.
First, A bad example
This is a simplified example of the wrong way to structure your app.
import React, { useState } from "react"
import axios from "axios"
import { useQuery } from "react-query"
interface Pokemon {
name: string
url: string
}
function PokemonApp() {
const [search, setSearch] = useState("")
const { data: pokemon, isLoading } = useQuery<Pokemon[]>(
"pokemon",
async () => {
const { data } = await axios.get(
"https://pokeapi.co/api/v2/pokemon?limit=100"
)
return data.results
}
)
const handleChange = (event: React.ChangeEvent<HTMLInputElement>) => {
setSearch(event.target.value)
}
if (isLoading) return <div>Loading...</div>
const filteredPokemon = pokemon.filter((p) => p.name.includes(search))
return (
<div>
<input type="text" value={search} onChange={handleChange} />
<ul>
{filteredPokemon.map((pokemon) => (
<li key={pokemon.name}>{pokemon.name}</li>
))}
</ul>
</div>
)
}
export default PokemonApp
We built a fun little Pokemon App like a mini Pokemon encyclopedia.
When you open it, it calls an API and grabs a list of all Pokemon. It can also pick a specific Pokemon and show its details when you click one.
Users can also type in the name of their favorite Pokemon.
The component will then look through the entire list to find a match and show only those ones.
Why is this a bad example?
This code is an excellent example of putting all your eggs in one basket.
And we all know how that ends :)
Can I build a CLI app on top of this code? No, I probably can't. Our component is wearing too many hats.
It's like a chef cooking, serving, washing dishes, and sweeping the floor afterward. It's a recipe for disaster.
Difficult to test
The current setup makes it challenging to test our code.
To test the search functionality, we have to deal with data fetching. but we must render the UI to test the data fetching.
It's as if we're assessing a chef's cooking while testing his dishwashing ability.
Hard to debug
Debugging is a burden when we must deal with many concerns simultaneously.
Say we have a bug that prevents a Pokemon's name from rendering. The issue could be a lousy API call, a flawed filtering logic, or a typo in the component.
Pinpointing the root cause is like finding a needle in a haystack without clear separation. You'd need to run and debug the entire app to find the root cause.
When you separate your concerns, you can isolate the problem by checking each part.
This will lead to a shorter time to resolution, and way less of a headache.
Impossible to reuse and replace
Imagine a new feature request from your product manager: a separate page for detailed Pokemon information.
With our current setup, we can't reuse our existing code. Our options are either to duplicate the code or to refactor it.
Replacing a piece from our component will also be challenging. Tightly coupled code restricts us from modifying individual parts without affecting others.
A better example
Let's transform our Pokemon App by separating concerns, and creating manageable pieces with clear responsibilities.
Splitting data fetching from UI
The first step is separating data-fetching from UI.
This lets us independently change our data source or UI and mock it for testing.
const fetchPokemons = () => {
const { data } = await axios.get('https://pokeapi.co/api/v2/pokemon?limit=100');
return data.results as Array<Pokemon>;
}
const usePokemons = () => {
return useQuery('pokemon', fetchPokemons);
};
Now, fetchPokemon fetches the data, and PokemonList displays it.
We can test fetchPokemon without worrying about the UI and vice versa.
We can even put a breakpoint on fetchPokemon to see what data it returns without running the app.
Decoupling the search functionality
We'll also separate the search functionality into a function.
const filterPokemons = (pokemon: Pokemon[], search: string) => {
return pokemon.filter((p) => p.name.includes(search))
}
// We can now use this logic anywhere we want
const allPokemon = await fetchPokemon()
const foundPokemon = searchPokemon(allPokemon, "pika")
Updating PokemonApp Component
With the separate components and functions, our main PokemonApp component becomes much cleaner:
function PokemonApp() {
const [search, setSearch] = useState("")
const { data: pokemon, isLoading } = usePokemon()
const handleChange = (event: React.ChangeEvent<HTMLInputElement>) => {
setSearch(event.target.value)
}
if (isLoading) return <div>Loading...</div>
const filteredPokemon = useFilteredPokemon(pokemon, search)
return (
<div>
<input type="text" value={search} onChange={handleChange} />
<PokemonList pokemon={filteredPokemon} />
</div>
)
}
We're only concerned with managing state and user interactions in PokemonApp, delegating the data fetching to fetchPokemon, filtering to searchPokemon, and UI display to PokemonList and Pokemon.
Testing
We can now test each piece of our app independently.
test("fetchPokemons", async () => {
const data = await fetchPokemons()
expect(data.length).toBe(100)
})
test("searchPokemon", () => {
const pokemon = [
{ name: "pikachu" },
{ name: "bulbasaur" },
{ name: "charmander" },
]
const filteredPokemon = searchPokemon(pokemon, "pika")
expect(filteredPokemon.length).toBe(1)
expect(filteredPokemon[0].name).toBe("pikachu")
})
Reusability
Let's build a CLI app on top of our logic.
const inquirer = require("inquirer")
async function run() {
const allPokemon = await fetchPokemon()
const answers = await inquirer.prompt([
{
type: "input",
name: "pokemonName",
message: "Which Pokemon are you looking for?",
},
])
const foundPokemon = filterPokemons(allPokemon, answers.pokemonName)
console.log(foundPokemon)
}
run()
Our decoupled design enables easy code reuse in different contexts.
We can now use the same logic to build another web, mobile, desktop, or even server app - anywhere that javascript allows us to run.
Conclusion
We used a simple example, but this concept applies to applications with thousands of components and functions.
Ultimately, our aim should be to construct applications where each piece has its defined role.
Your code can be a well-organized kitchen:
the chef, the waitperson, and the dishwasher have unique responsibilities.
Think about it when you're about to write your next web application:
Can I re-use the code to create a CLI tool without changing much of the code?
--
Thanks for reading!
You can find me on Github, Twitter, or my website.
Feel free to reach out :)
Top comments (0)