Lint is incorrect for constructor : Inferno Error: cannot update state via setState() in constructor. Instead, assign to `this.state` directly or define a `state = {};`

[dessalines]

If I use state = ... in a constructor, it fails lint. If I follow lint and use setState(, I get the following error:

404: Error: Inferno Error: cannot update state via setState() in constructor. Instead, assign to this.state directly or define a state = {};

[Havunen]

If I use state = ... in a constructor, it fails lint.

I think the linter is trying to say that it might cause unexpected behavior if you assign state directly. Because then any derived class by assigning the state in constructor would override the assigned value done in base class.

This is expected behavior in Inferno and in React too.

One option is to set the properties to state in constructor:

class MyComponent extends Component {
  constructor(props) {
    super(props)
    this.state.counter = 0;
  }
}

or by using object assign

class MyComponent extends Component {
  constructor(props) {
    super(props)
    
    Object.assign(this.state, {
        counter: 0
    });
  }
}

[dessalines]

this.state.counter = 0;

This would fail lint. I'm getting the following message when I do that:

Cannot assign to 'postRes' because it is a read-only property. 
131       this.state.postRes = Some(this.isoData.routeData[0] as GetPostResponse);

I'd also much prefer not using Object.assign because it loses all my type info.

[Havunen]

ah yes, I will fix that

[Havunen]

inferno 8.0.2 should allow setting properties directly to state infernojs/inferno@d0ccb38

[dessalines]

This is still an issue when using closures in the constructor. For example, try :

setTimeout(() => this.state.counter = 0, 10);

in the constructor.

[dessalines]

Just posting this because you might find it useful, but I have to use closures because I recently converted all typescript style optionals ( IE ? ) , to rust-style Option and Result using this library: https://github.com/sniptt-official/monads . Then you can handle the nonexistent cases really easily with match:

myOptional.match({
  some: val => // Do something
  none: // Do something else
});

It eliminated a whole class of NullPointerException type bugs, and was really useful especially for API-type libraries.

[dessalines]

Another thing: would it be possible to have a lint that dissallows using setState in the constructor? Otherwise these bugs become runtime errors rather than compile-time ones.

If not, or if that should be a different issue, that's no problem.

[Havunen]

Yeah, thats good idea. Can you create a new issue please 👍

[Havunen]

If I use state = ... in a constructor, it fails lint

What linting rule is that? I think my previous fix should be reverted, I somehow thought readonly made that, but of course readonly property can be set in constructor

[dessalines]

This one: https://github.com/infernojs/eslint-plugin-inferno/blob/master/docs/rules/no-direct-mutation-state.md

And it seems to only be violated if I use it in a closure in the constructor.

[Havunen]

Ah, just normal closure in constructor? That sounds like a bug in the eslint rule. Can you post a small code example just to be sure! Thanks!