etienney

I'm also very interested as it is similar to my problem [#447](https://github.com/patrick-kidger/diffrax/issues/447) It is not the most satisfying solutio because : + doing the computation like that is probably computationaly...

> So my understanding is that these kinds of hybrid systems usually have a particular time at which the discrete mode switches? This means that a solver-based step-by-step probably isn't...

> 0.5.1 is pre the events refactor right? Yes it is ! ( Though I'm still using this version for one main reason that is state.tnext is not defined in...

> As far as I can tell, this happens exactly because your condition depends on `tnext`. This should be possible right now in `0.6.0` (at least not in a straight-forward...

Yes, this is indeed the obvious and perfect solution for me ! But as I said this is not possible. The "real" y (not aux) i'm integrating sould be integrated...

It is a little bit of both, ie I'm using an adaptative step size, but the problem does not come from the step size controller struggling due to aux not...

Okay `save_state `may be too much to be given to users but maybe save_index, save_state.ts and save_state.ys would be nice ? The usecase is to compute functions with entries like...

Then you would not be able to use your saved output in some event though. Of course this can be done at the end, but the idea is to do...

> Is there a way to have access to y_dot inside of SaveAt ? Maybe what you want to do is putting the option "steps=True" to sync the values at...

Thanks you both for your answers ! > I'm not sure why you want to use Euler instead of Tsit5's embedded error estimate, though. Using the latter will actually be...