Installation fails with current MLR version

[tbrieden]

As described in the installation section, the installation of automlr is currently only possible by using the adapted MLR version that contains a specific CPO extension. However, this version of MLR is not up-to-date and the pull request mentioned in the installation section is already merged into MLR. Therefore, I was trying to install automlr with the current MLR version(2.12.1). This fails with the following error message:

roxygenise()
First time using roxygen2. Upgrading automatically...
Updating roxygen version in /tmp/RtmpYSmN2Q/devtoolsf733f14c882/mlr-org-automlr-8e74e08/DESCRIPTION
Writing NAMESPACE
Loading automlr
Loading required package: mlr
Loading required package: ParamHelpers
Error in makeCPO("cfimputer", .datasplit = "factor", .properties.adding = "missings", (from CPO.R#24) :
could not find function "makeCPO"

It suspect that the makeCPO method was removed again from MLR and moved to mlrCPO. Therefore, I installed mlrCPO and added it as a dependency to a local clone of the automlr repo. Unfortunately, the installations fails again, giving me the following error:

roxygenise()
First time using roxygen2. Upgrading automatically...
Updating roxygen version in /tmp/RtmpYSmN2Q/filef735918e0f6/automlr/DESCRIPTION
Writing NAMESPACE
Loading automlr
Loading required package: mlr
Loading required package: ParamHelpers
Loading required package: mlrCPO
Error in makeCPO("cfimputer", .datasplit = "factor", .properties.adding = "missings", (from CPO.R#24) :
unused arguments (.datasplit = "factor", .properties.adding = "missings", .stateless = TRUE, cpo.trafo = {
for (n in names(data)) {
data[[n]] = factor(data[[n]], levels = c("MISSING", levels(data[[n]])))
data[[n]][is.na(data[[n]])] = "MISSING"
}
data
})

It seems that the CPO calls used in master/R/CPO.R are outdated. It there any chance to get automlr running with the current MLR release or would this require a lot of work on automlr to get this up-to-date?

[tbrieden]

I switched to using mlrMBO directly instead of automlr. Therefore, the questions I asked here are not that important to me anymore.