DEV Community

Giancarlo Buomprisco for Angular

Posted on • Updated on • Originally published at angularbites.com

Angular: Setters vs ngOnChanges - which one is better?

Alt Text

This post was originally published on Angular Bites

Getting notified about an Angular component's property changes is normally done in 2 ways:

  • adding a setter to the property
  • using the ngOnChanges lifecycle hook

But... is there a best practice?

This discussion recently came up with my colleagues while trying to establish a standard practice in our codebase. We tried to find objective arguments to understand which one is better.

As usual, the answer depends on the scenario.

Style

Style is very much a subjective factor, but using a setter is hands-down my favorite approach. Let's take a look at a common scenario:

class MyComponent {
  private subject$ = new Subject<string>();

  @Input()
  set name(name: string) {
    this.subject$.next(name);
  }
}

It's succinct, type-safe, and encourages the usage of Observables. Not much to dislike, imho.

But can you not add a getter?

Yes. It turns out, Angular does not check the previous value by invoking the getter on the property, but stores its value in its component's logical view.

If you're interested in reading the source code where this happens, check this out.

class MyComponent implements OnChanges {
  @Input() name: string;

  private subject$ = new Subject<string>();

  ngOnChanges(changes: SimpleChanges) {
    // changes.name.currentValue is typed as `any`
    this.subject$.next(changes.name.currentValue);
  }
}

The ngOnChanges lifecycle hook, on the contrary, it's not as nice (in my opinion) - and most importantly, is weakly typed.

Also - it's worth to mention that using setters usually takes less code, which is always a good thing.

Performance

Does performance change much? At first, we thought that ngOnChanges would be more efficient as being part of Angular's lifecycle hooks, and therefore being aware of when a property changed.

It turns out, though, that Angular does only change a property when the binding is a new instance. Of course, we're taking into account the change detection being OnPush.

Performance-wise, according to my tests, there isn't a better way, and shouldn't be a factor when deciding which way to go with.

Dealing with multiple Inputs

The situation changes when taking into account changes on multiple inputs:

class MyComponent implements OnChanges {
  @Input() name: string;
  @Input() email: string;

  private username$ = new Subject<string>();

  ngOnChanges({ name, email }: SimpleChanges) {
    const username = name.currentValue || email.currentValue;
    this.username$.next(username);
  }
}

In this case, it's fairly straightforward and simpler to receive all the inputs at once.

But because this situation is pretty uncommon, and sometimes a sign of a code-smell, you'll find yourselves wanting to use the setter the majority of the time.

At the end of the day, remember that this decision is always up to you and your team's preferences.

Hope you enjoyed this post. Ciao!

Discussion (8)

Collapse
bo profile image
Bo Vandersteene
  ngOnChanges({ name, email }: SimpleChanges) {
    const username = name.currentValue || email.currentValue;
    this.username$.next(username);
  } 

If you take this code in consideration, you need to pay attention, that both values are changed on the same time. If the ngOnChanges is called 2 times for each value, than this one can give some unwanted sideEffects.

Collapse
gc_psk profile image
Giancarlo Buomprisco Author

Hi! do you mean that name or email could be undefined? TBH I'd never ship this to production - just needed a short example with multiple inputs ๐Ÿ˜…

Collapse
bo profile image
Bo Vandersteene • Edited

Ok so if you have a component that uses your code like this

<my-component [email]="email$ | async" 
                               [username]="username$ | async"> </my-component>

If the input of the observables look like this

email$    = -E----
username$  =  ----U--

Then the ngOnChanges will be called first with only email and then with only username so one of the two is undefined. Only the values that are changed are submitted in the SimpleChanges

Collapse
hassam7 profile image
Hassam Ali

I tend to avoid setters with the inputs if the inputs have dependency on other inputs because then the order inputs to the components matter.

@Component({...})
class Component {
  @Input()
  url: Url;
  @Input()
  set users(users: Array<User>) {
    this._users = users.map((user) => {
      return user.avatar = this.url.baseUrl + user.avatar;
    });
  }
  private _users: Array<User>;
}
<component [url]="'someurl'" [users]=users /> //will work fine
<component [users]=users [url]="'someurl'" /> //will not work fine
Enter fullscreen mode Exit fullscreen mode

see this link for further details: medium.com/generic-ui/a-deep-dive-...

Collapse
glebirovich profile image
Gleb Irovich

Surprisingly we had this discussion as well recently. We came up with using setters as a preferred option. I donโ€™t see any major disadvantages of that approach. ๐Ÿ‘

Collapse
constjs profile image
Piotr Lewandowski

Great and short summary :)
Have you considered performance differences with default change detection strategy?

Collapse
gc_psk profile image
Giancarlo Buomprisco Author • Edited

I haven't - but even then I'd be surprised if there were any noticeable differences :)

Collapse
penguintheorem profile image
Attilio Urbani

Hi!

i prefer the first approach too (setter). I think that the code is definitely more clean than using the hook. Thanks for sharing your tip!