Default measure in $aggregate() should use the available predict_set (and not always "test", irrespective of whether test predictions exist) #1303
Description
The default configuration for $aggregate() should use the available predict set (if there is exactly one predict set available, as is the case in the example below).
library(mlr3)
task = tsk("iris")
task$internal_valid_task = sample(1:150, 50)
learner = lrn("classif.featureless", predict_sets = "internal_valid")
rr = resample(task, learner, rsmp("insample"))
#> INFO [10:25:22.780] [mlr3] Applying learner 'classif.featureless' on task 'iris' (iter 1/1)
rr$aggregate()
#> Warning in assert_measure(measure, task = task, learner = learner, prediction =
#> prediction, : Measure 'classif.ce' needs predict sets 'test', but learner
#> 'classif.featureless' only predicted on sets 'internal_valid'
#> classif.ce
#> NaNCreated on 2025-05-06 with reprex v2.1.1