Actions without types throw an error rather than return the default state

[JaKXz]

In my test setup code I have this line:

test.beforeEach(t => {
  const initialState = reducer();
});

where

const reducer = handleActions({ ... }, { /*initial state*/});

However, the first method throws an error because:

and in the transpiled source it looks like action.type.toString() is called unguardedly. Now, maybe I'm misunderstanding how FSAs work, but I think it makes sense for a reducer test to be able to check that the initial state is returned when the action type is missing. I can try to make a PR for this as well, if you think I'm correct about the expected behaviour.

If not, could you please fill me in on what I'm missing? Thanks!

[yangmillstheory]

A reducer returns the default state when the state given is undefined; the action must exist. This isn't FSA thing, just a Redux thing.

Your test is throwing an expected error because you haven't given it an action; reducers respond to actions.

For tests around this, see

• https://github.com/acdlite/redux-actions/blob/master/src/__tests__/handleAction-test.js#L170
• https://github.com/acdlite/redux-actions/blob/master/src/__tests__/handleAction-test.js#L15

For background material, see

• Confusion about initial state, reducers and the Store. reduxjs/redux#1189
• Why does a reducer needs a default state? reduxjs/redux#514

[JaKXz]

Thanks @yangmillstheory - the background material helps.

Follow-up question:

I actually originally tried reducer(undefined, {}) but that failed with the same error. Could that be a valid set of arguments for handleAction?

[yangmillstheory]

See #139 (comment) and the comment below it.

[JaKXz]

Sorry for the duplication/noise, and 🙏 for a great project! I thought I'd searched the issues properly but I guess not.

[yangmillstheory]

Out of curiosity, when do you encounter actions in your app that don't have a type property?

If you're using this library, createActions makes action creators that will always return actions with type.

[JaKXz]

Yeah at this point it's just 🚴🏚-ing about test setup code, I can't imagine a real example of having an undefined action or action type being dispatched. And I think I understand the argument that having the "robustness" to handle that is more bad than good.