discrete parameters with trafo break printing of log messages

[schiffner]

See below.
In discreteValueToName transformed and original values are compared, which generates the error.

library(mlr)
options(error = recover)

ps = makeParamSet(
	makeDiscreteParam("cost", values = 1, trafo = function(x) 2^x)
)
tuneParams("classif.svm", iris.task, resampling = cv3, control = makeTuneControlGrid(), par.set = ps)
# [Tune] Started tuning learner classif.svm for parameter set:
         # Type len Def Constr Req Tunable Trafo
# cost discrete   -   -      1   -    TRUE     Y
# With control class: TuneControlGrid
# Imputation value: 1
# Error in getIndex(par$values, x) : Value not found!

# Enter a frame number, or 0 to exit   

 # 1: tuneParams("classif.svm", iris.task, resampling = cv3, control = makeTuneControlGrid(), par.set = ps)
 # 2: sel.func(learner, task, resampling, measures, par.set, control, opt.path, show.info, resample.fun)
 # 3: evalOptimizationStatesTune(learner, task, resampling, measures, par.set, control, opt.path, show.info, xs, dobs = seq_along(xs), eols = NA, remov
 # 4: evalOptimizationStates(learner, task, resampling, measures, par.set, NULL, control, opt.path, show.info, states, dobs, eols, remove.nas, resample
 # 5: parallelMap(evalOptimizationState, dobs, states, level = level, more.args = list(learner = learner, task = task, resampling = resampling, measure
 # 6: mapply(fun2, ..., MoreArgs = more.args, SIMPLIFY = FALSE, USE.NAMES = FALSE)
 # 7: (function (learner, task, resampling, measures, par.set, bits.to.features, control, opt.path, show.info, dob, state, remove.nas, resample.fun) 
# {
 # 8: log.fun(learner, task, resampling, measures, par.set, control, opt.path, dob, state, NA, remove.nas, stage = 1)
 # 9: paramValueToString(par.set, x, show.missing.values = !remove.nas)
# 10: paramValueToString.ParamSet(par.set, x, show.missing.values = !remove.nas)
# 11: sprintf("%s=%s", pn, paramValueToString(p, val, show.missing.values, num.format))
# 12: paramValueToString(p, val, show.missing.values, num.format)
# 13: paramValueToString.Param(p, val, show.missing.values, num.format)
# 14: discreteValueToName(par, x)
# 15: getIndex(par$values, x)

# Selection: 14
# Called from: top level 
# Browse[1]> ls()
# [1] "getIndex" "ns"       "par"      "x"       
# Browse[1]> par$values
# $`1`
# [1] 1

# Browse[1]> x
# [1] 2

[berndbischl]

@schiffner why would you ever do that? its a discrete param. just enumerate the values you want.
logscale or normal.

if can for sure simply remove the trafo arg for these params?

[schiffner]

@schiffner why would you ever do that?

Because I can?! And the docs say I can?! And no one stopped me... :)
It "happened" to me because I switched between tuning methods from random to grid search and had used the trafo arg earlier when the param was still numeric ...

I'm also ok with removing the trafo arg.

[berndbischl]

Because I can?! And the docs say I can?! And no one stopped me... :)

just trying to figure out whether it makes sense to remove the trafo.....
@mllg ?

[berndbischl]

@larskotthoff