sl3_Task: proper handling of outcome=NULL or allow missing outcome about sl3 HOT 6 CLOSED

Somewhat addressed in #72. task$Y for NULL outcome generates an error:

> task <- sl3_Task$new(cpp, covariates = covars, outcome = NULL)
> task$Y
Error in generator_fun(node_name, self$nrow) : Node outcome not specified

I kind of like having to force outcome=NULL if you really want a task without an outcome. Do you think it's important to be able to allowing missing(outcome)?

from sl3.

I think so. I think it might be relevant for prediction, which I imagine would be a typical use-case of the package. So I feed a trained sl3 learner a bunch of new data, but I now have to remember each time to set outcome = NULL? I feel like its too much to ask of a user to remember. On the other hand, is it difficult to add missing(outcome) check that sets it to NULL?

from sl3.

So, I am OK with the above error, it makes sense to me. I guess all we need is a check for missing(outcome)?

from sl3.

No it's of course super easy to add that functionality. I just thought we might want to provide some signal to a user that they didn't specify an outcome so they're not surprised later. If you don't think that's necessary I'll go ahead and support missing outcome

from sl3.

Awesome! Thanks. I think this is fine.

from sl3.

Fixed in #76

from sl3.