What is the correct way to make a discrete integer param?

[jakob-r]

What is the best way to achieve this in ParamHelpers?

ps = makeParamSet(makeDiscreteParam("a", c(1L,10L,25L)), makeDiscreteParam("b", letters[1:2]))
des = generateGridDesign(ps, resolution = 1)
class(des$a)
# [1] "factor"
des$a = as.integer(as.character(des$a))
des

[jakob-r]

2 do: Integrate: generateGridDesign(ps, discrete.as.values = FALSE)
as well as converter function for data.frames which uses discreteNameToValue

[berndbischl]

Note that 3 genDesign functions already exist

[jakob-r]

It just came to my mind that discrete.as.values == TRUE would just involve the call of convertListOfRowsToDataFrame(dfRowsToList(des, par.set = ps)) this is on the one hand super easy so we should leave it out and let the user do it but on the other hand it does not look really sexy for the user.
So in the end we might just have a helper function? Which then again might be forgotten when a user stumbles upon the problem.

[berndbischl]

The problem is that the "values" in the discrete param can be non-scalar.
It is still possible to put them into a data.frame, but some users might be "weirded out" by this.
It is still probably
a) consistent
b) super useful if your discrete are, as often, just scalars.

[jakob-r]

I would propose to throw an error if non-scalar values are in the given design. Nobody sane would actually want to use that helper with non-scalar values anyhow I guess.