Gil

> Really good :) > I made a round of comments ... no rush to address them, but would be good to do that before we merge this. What else...

> What are 9602 and 4423? Those are the IDs of a flow and a task that I used with the openml-python extension I wrote and i know they work

Yup, that’s correct :) so you would probably have to return something like shared_from_this() in put

Hmm, try to get it to run and send a PR and then we can have a look :)

> That is no problem, we can expose that lazily via `get` and functions that take no arguments (I think). The issue is that those functions require `Labels` and as...

> ah! crap! :) > TBH I don't think it is such a problem.... I would actually just not expose it for now. > But these cases require a solution...

> they are computed anyways when fitting the model. > But how would you compute them in apply? Would you store them? Dont think that is any different from storing...

@karlnapf I guess the idea is that these values are observed and you emit the observation when calling `train`/`apply`? I guess then it would be the values that are exposed,...

I guess that means every arithmetic type can be casted to any other arithmetic type?

It depends. With -Wconversion you get a warning and I guess if you combined that with -Werror you get a compiler error. Without that warning the floating point type is...