Move pkg function calls within ParamSets to helper file (mlr-org#2730)

* Refactor function calls from packages (`<pkg::fun>`) within ParamSets (mlr-org#2730)

Author: pat-s

R/RLearner_classif_mda.R

@@ -11,8 +11,7 @@ makeRLearner.classif.mda = function() {
       makeIntegerLearnerParam(id = "dimension", lower = 1L),
       makeNumericLearnerParam(id = "eps", default = .Machine$double.eps, lower = 0),
       makeIntegerLearnerParam(id = "iter", default = 5L, lower = 1L),
-      makeDiscreteLearnerParam(id = "method", default = mda::polyreg,
-        values = list(polyreg = mda::polyreg, mars = mda::mars, bruto = mda::bruto, gen.ridge = mda::gen.ridge)),
+      makeDiscreteLearnerParam(id = "method", default = "polyreg", values = list("polyreg", "mars", "bruto", "gen.ridge")),
       makeLogicalLearnerParam(id = "keep.fitted", default = TRUE),
       makeLogicalLearnerParam(id = "trace", default = FALSE, tunable = FALSE),
       makeDiscreteLearnerParam(id = "start.method", default = "kmeans", values = c("kmeans", "lvq")),
@@ -29,9 +28,18 @@ makeRLearner.classif.mda = function() {
 }
 
 #' @export
-trainLearner.classif.mda = function(.learner, .task, .subset, .weights = NULL, ...) {
+trainLearner.classif.mda = function(.learner, .task, .subset, .weights = NULL, method, ...) {
   f = getTaskFormula(.task)
-  mda::mda(f, data = getTaskData(.task, .subset), ...)
+  args = list(...)
+  if (!missing(method)) {
+    if (is.character(methods)) {
+      args$method = getFromNamespace(method, "mda")
+    } else {
+      args$method = method #this allows to set the method if on.par.out.of.bounds is set to "warn" or "quiet"
+    }
+  }
+  args = c(list(f, data = getTaskData(.task, .subset)), args)
+  do.call(mda::mda, args)
 }
 
 #' @export

R/RLearner_regr_mob.R

@@ -10,22 +10,21 @@ makeRLearner.regr.mob = function() {
       makeNumericLearnerParam(id = "trim", default = 0.1, lower = 0, upper = 1),
       makeLogicalLearnerParam(id = "breakties", default = FALSE),
       makeLogicalLearnerParam(id = "verbose", default = FALSE, tunable = FALSE),
-      makeDiscreteLearnerParam(id = "model", default = modeltools::glinearModel,
-        values = list(glinearModel = modeltools::glinearModel, linearModel = modeltools::linearModel)),
+      makeDiscreteLearnerParam(id = "model", default = "glinearModel", values = list("glinearModel", "linearModel")),
       makeUntypedLearnerParam(id = "part.feats"),
       makeUntypedLearnerParam(id = "term.feats")
     ),
     par.vals = list(),
     properties = c("numerics", "factors", "weights"),
     name = "Model-based Recursive Partitioning  Yielding a Tree with Fitted Models Associated with each Terminal Node",
     short.name = "mob",
-    callees = c("mob", "mob_control", "glinearModel", "linearModel")
+    callees = c("mob", "mob_control")
   )
 }
 
 #' @export
 trainLearner.regr.mob = function(.learner, .task, .subset, .weights = NULL, alpha, bonferroni, minsplit,
-  trim, breakties, verbose, part.feats, term.feats, ...) {
+  trim, breakties, verbose, part.feats, term.feats, model, ...) {
 
   cntrl = learnerArgsToControl(party::mob_control, alpha, bonferroni, minsplit, trim, breakties, verbose)
 
@@ -42,11 +41,20 @@ trainLearner.regr.mob = function(.learner, .task, .subset, .weights = NULL, alph
   target = getTaskTargetNames(.task)
   f = as.formula(stri_paste(target, "~", collapse(term.feats, sep = " + "), "|", collapse(part.feats, sep = " + "), sep = " "))
 
-  if (is.null(.weights)) {
-    model = party::mob(f, data = getTaskData(.task, .subset), control = cntrl, ...)
-  } else {
-    model = party::mob(f, data = getTaskData(.task, .subset), control = cntrl, weights = .weights, ...)
+  args = list(f, data = getTaskData(.task, .subset), control = cntrl, ...)
+  if (!is.null(.weights)) {
+    args$weights = .weights
   }
+  if (!missing(model)) {
+    if (is.character(model)) {
+      args$model = getFromNamespace(model, "mda")
+    } else {
+      args$model = model #this allows to set the model if on.par.out.of.bounds is set to "warn" or "quiet"
+    }
+  }
+
+  model = do.call(party::mob, args)
+
   # sometimes mob fails to fit a model but does not signal an exception.
   if (anyMissing(coef(model))) {
     stop("Failed to fit party::mob. Some coefficients are estimated as NA")