Naming runs, prefixing predictions

[Peder2911]

Runs should have a required name attribute. This attribute should then be applied to predictions created in the run.

[hhegre]

While revising this; I don't think it is necessary to store all the predictors in the output. But possibly it is useful to store the prediction target as a column, with the same prefix to avoid overwriting.

[Peder2911]

While it might not be necessary, it may be much less complicated to just store everything rather than only storing something conditionally. Storage is very cheap, and complexity is not. Agreed?

Storing the outcome might be a good idea. While it is available elsewhere, storing the outcome that was used at the time of running is good for reproducibility.

Input @mihaicroicu ?

[hhegre]

I think a constraint is that using the viewser database means that column names has to be distinct to avoid overwriting. Besides, given the step shifting I don't see predictors contemporaneous with non-shifted outcome (which I presume this is) is not so useful? Given that, storing adds complexity since users would want to subset after reading. The outcome is different.

[hhegre]

I am not sure it is useful that the predictors come out of the step shifting /views runs procedure either

[Peder2911]

Not sure I understand what direction we went with this. From the issue, the idea is just to have ViewsRun output predictions that are prefixed with a name, which is a required attribute of ViewsRun on instantiation?

[Peder2911]

Update @hhegre ?

[Peder2911]

Unassigned as I am leaving the project.