SET_SUBMIT_SUCCEEDED called from empty onSubmit and after stopSubmit with errors

I’m not returning a Promise from my onSubmit, instead dispatching an action to be later picked up in a saga.

The saga calls, startSubmit('formName'), and then at some point later calling stopSubmit('formName', errors) (or just stopSubmit('formName') if everything was ok).

In the case of a failure, the errors are being properly picked up and applied to the form, however, I’m seeing the @@redux-form/SET_SUBMIT_SUCCEEDED action coming through as well, which results in an inconsistent form state. The submit failed, but @@redux-form/SET_SUBMIT_SUCCEEDED is setting submitSucceeded to true even though I signalled a failed submission.

I’m trying to signal a failure, but the action sequence I’m getting is,

What’s more, I get the @@redux-form/SET_SUBMIT_SUCCEEDED action creeping in even if I do nothing at all in onSubmit, i.e. with an empty onSubmit function that does nothing and doesn’t even return a Promise I see the following action,

In terms of correctness, correct me if I’m wrong, but I feel like this action shouldn’t be being called after an empty onSubmit that doesn’t return a Promise. In that case shouldn’t nothing have happened? A Promise wasn’t returned, and startSubmit() wasn’t called so why is Redux Form flipping submitSucceeded to true?

[erikras]

Proposal

A new function, like handleSubmit, called dispatchSubmit, that you use like this:

<form onSubmit={this.props.dispatchSubmit}>
  ...
</form>

It will check that sync validation and async validation all passes, and if and only if the form is valid, it will dispatch an action that looks something like:

{
  type: '@@redux-form/DISPATCH_SUBMIT', // 'SUBMIT' is taken already... 😕
  payload: { 
    // all your form values
  }
}

The redux-form reducer will ignore this action entirely. It will be up to all your fancy async middlewares to dispatch startSubmit(), stopSubmit(), etc. to update the form’s submitting state manually.

Would that satisfy everyone’s needs?

[ecnaidar]

@erikras This looks interesting, though still feels a lot like hack and adds complexity. If there is still interest, I might get a PR going for the dispatchSubmit.