saga builder does not allow accessing logic fields

[rschyboll]

Hi, it's me, again :D

Another thing i noticed when using the kea-saga plugin, the saga builder, which allows starting the saga after mounting the logic, does not accept as input a callback, to allow access logic parameters, like props for example.

I do see why this is the case, as it is impossible to differentiate a saga between a callback which accepts a logic, as both are functions in js.

A proposal that i would have, in my own project I've written such a builder:

export function startSagas<L extends Logic = Logic>(
  input: (logic: L) => Saga[] | Saga,
): LogicBuilder<L> {
  return (_logic) => {
    const logic = _logic as Logic as LogicWithSaga;
    addGetAndFetch(logic);

    const sagas = input(_logic);

    if (Array.isArray(sagas)) {
      for (const localsaga of sagas) {
        saga(localsaga)(_logic);
      }
    } else {
      saga(sagas)(_logic);
    }
  };

It's just a builder that allows launching one or multiple sagas after logic mounted, with access to all logic fields. Maybe adding such a builder to the plugin would be helpful for other people, if you want I could make a pull request with it.

[mariusandra]

Hey, in a saga you can normally use this to access the logic itself. E.g. this.actions.doBla(). We use function.bind(logic) in many places in the builder.

Would that work for you as well?

[rschyboll]

Hey, it does only partially solve my issue, as I'm writing with typescript, and using this, does not provide any type annotations :( Is there maybe some other solution to this?

[mariusandra]

Hmm... one other option is to just use the logic itself:

const someLogic = kea([
  saga(() => {
    console.log(someLogic.values.bla)
    const { doSomething } = someLogic(this.props)
  })
])

However, I noticed that all the sagas just get zero arguments, so I think there's really no reason to just not pass on the logic as the first argument. So 3.1.0 does that.

It'll be a breaking change to whoever is reusing a saga/worker with a default first argument directly in a logic, but I think it's an acceptable risk.