Fix bug where predicting on regression forest with no newData caused

crash.

R/predict.R

@@ -99,7 +99,7 @@ predict.JRandomForest <- function(object, newData=NULL, parallel=TRUE, out.of.ba
   predictionsJava <- .jcall(forestObject, makeResponse(.class_List), function.to.use, predictionDataList)
 
   if(predictionClass == "numeric"){
-    predictions <- vector(length=nrow(newData), mode="numeric")
+    predictions <- vector(length=numRows, mode="numeric")
   }
   else{
     predictions <- list()

tests/testthat/test_running.R

@@ -28,6 +28,7 @@ test_that("Regresssion doesn't crash", {
   forest <- train(y ~ x, trainingData, ntree=50, numberOfSplits=0, mtry=1, nodeSize=5, cores=2, displayProgress=FALSE)
   
   predictions <- predict(forest, testData)
+  other_predictions <- predict(forest) # there was a bug if newData wasn't provided.
   
   expect_true(T) # show Ok if we got this far