Add also List hyper-parameter
Currently we have a Set hyper-parameter which has an interesting property that the order still matters to assure reproducibility.
But it seems there are also two other cases to consider:
- A case where order matters for other reasons, for example, a list of dataframes to concatenate (not that this would be a good hyper-parameter, probably it should be just a regular argument/input).
- Set implies that the values should not be repeated, but in some cases it is OK that they are. For example, a list of some constants as coefficients. It would be tricky to require that no coefficient value should ever repeat.
So, maybe we should add an explicit hyper-parameter class List where it is clear that values can be repeated (and is sampled in this way).
Now, should we add also OrderedSet or something like that? The point is that Set is already ordered, but that is so that outputs are reproducible. So one can configure/decide when comparing two pipelines for equality, to compare them with respecting the order in Set hyper-parameters or not.