fix and finish expression handling

[berndbischl]

expr handling was merged, but is wrong.

PR was here:
#1126

revert was done here:
981bc9b

the dif shows the code that would need to be reimplemented, and then fixed of course.

analysis:

1. this example shows it does not work

load_all()
task = makeClassifTask(data = iris, target = "Species")
dict = getTaskDictionary(task = task)
lrn1 = makeLearner("classif.rpart", minsplit = expression(n))
print(getHyperPars(lrn1))
m = train(lrn1, task)
print(m$learner.model$control$minsplit)
m = train(lrn1, task, subset = 1:100)
print(m$learner.model$control$minsplit # is 150! wrong

2. that such a basic thing as in 1) was not unit tested in the orig PR seems to be a problem in itself.

3. this is due to the .subset arg in trainLearner. i remember it had a reason that we do the subsetting there, but this seems quite error prone, we also had problems in other places. maybe we remove this? this must be discussed in another issue and should not be tight-coupled (hopefully?) with resolving our problem here.

[jakob-r]

1 + 2. See my answer in the PR. This behavior was never specified nor is it "falsely advertised". Even though it makes sense I don't see any "faulty behavior" just a requested feature that was missing. We can now discuss if n should be the Task size or the size of the training set. I would propose to have the following dictionary entries: n, train.n and predict.n.

[berndbischl]

Can we agree it is counter intuitive and not what we want? Can you give an example where what I posted in my example is intended behavior?

And I am pretty sure how you now specify learner defaults is really wrong

[berndbischl]

Just assume this. A learner has arg x. The package docs say: x is by default n/2. What do you write in the mlr learner default expression?

[berndbischl]

And i don't see the use of all of these symbols you suggest? I mean n should refer exactly to what the learner sees? For the learner there is no task?

[berndbischl]

Just assume this. A learner has arg x. The package docs say: x is by default n/2. What do you write in the mlr learner default expression?

[jakob-r]

New PR here: #1576

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.