Fix/predict newdata (#1240)

* ci: fail on note

* fix(predict): type conversion when predicting on new data

* ...

---------

Co-authored-by: Marc Becker <[email protected]>

NEWS.md

@@ -1,7 +1,9 @@
 # mlr3 (development version)
 
+* fix: the `$predict_newdata()` method of `Learner` now automatically conducts type conversions if the input is a `data.frame` (#685)
+* BREAKING_CHANGE: Predicting on a `task` with the wrong column information is now an error and not a warning. 
 * Column names with UTF-8 characters are now allowed by default.
-The option `mlr3.allow_utf8_names` is removed.
+  The option `mlr3.allow_utf8_names` is removed.
 * BREAKING CHANGE: `Learner$predict_types` is read-only now.
 * docs: Clear up behavior of `Learner$predict_type` after training.
 

R/Learner.R

@@ -377,6 +377,8 @@ Learner = R6Class("Learner",
     #'   `data.frame()` or [DataBackend].
     #'   If a [DataBackend] is provided as `newdata`, the row ids are preserved,
     #'   otherwise they are set to to the sequence `1:nrow(newdata)`.
+    #'  If the input is a `data.frame`, [`auto_convert`] is used for type-conversions to ensure compatability
+    #'  of features between `$train()` and `$predict()`.
     #'
     #' @param task ([Task]).
     #'
@@ -393,6 +395,14 @@ Learner = R6Class("Learner",
         task = task_rm_backend(task)
       }
 
+      if (is.data.frame(newdata)) {
+        keep_cols = intersect(names(newdata), task$col_info$id)
+        ci = task$col_info[list(keep_cols), on = "id"]
+        newdata = do.call(data.table, Map(auto_convert,
+          value = as.list(newdata)[ci$id],
+          id = ci$id, type = ci$type, levels = ci$levels))
+      }
+
       newdata = as_data_backend(newdata)
       assert_names(newdata$colnames, must.include = task$feature_names)
 
@@ -409,6 +419,7 @@ Learner = R6Class("Learner",
 
       # do some type conversions if necessary
       task$backend = newdata
+      task$col_info = col_info(task$backend)
       task$row_roles$use = task$backend$rownames
       self$predict(task)
     },

R/assertions.R

@@ -189,7 +189,7 @@ assert_predictable = function(task, learner) {
       all(pmap_lgl(list(x = ci_train$levels, y = ci_predict$levels), identical))
 
     if (!ok) {
-      lg$warn("Learner '%s' received task with different column info (feature type or level ordering) during train and predict.", learner$id)
+      stopf("Learner '%s' received task with different column info (feature type or factor level ordering) during train and predict.", learner$id)
     }
   }
 

tests/testthat/test_Learner.R

@@ -255,13 +255,15 @@ test_that("learner cannot be trained with TuneToken present", {
 
 test_that("integer<->numeric conversion in newdata (#533)", {
   data = data.table(y = runif(10), x = 1:10)
-  newdata = data.table(y = runif(10), x = 1:10 + 0.1)
+  newdata1 = data.table(y = runif(10), x = as.double(1:10))
+  newdata2 = data.table(y = runif(10), x = 1:10 + 0.1)
 
   task = TaskRegr$new("test", data, "y")
   learner = lrn("regr.featureless")
   learner$train(task)
   expect_prediction(learner$predict_newdata(data))
-  expect_prediction(learner$predict_newdata(newdata))
+  expect_prediction(learner$predict_newdata(newdata1))
+  expect_error(learner$predict_newdata(newdata2), "failed to convert from class 'numeric'")
 })
 
 test_that("weights", {
@@ -575,10 +577,7 @@ test_that("column info is compared during predict", {
   task_other = as_task_classif(dother, target = "y")
   l = lrn("classif.rpart")
   l$train(task)
-  old_threshold = lg$threshold
-  lg$set_threshold("warn")
-  expect_output(l$predict(task_flip), "task with different column info")
-  lg$set_threshold(old_threshold)
+  expect_error(l$predict(task_flip), "task with different column info")
   expect_error(l$predict(task_other), "with different columns")
 })
 
@@ -663,3 +662,19 @@ test_that("configure method works", {
   expect_equal(learner$param_set$values$xval, 10)
   expect_equal(learner$predict_sets, "train")
 })
+
+test_that("predict_newdata auto conversion (#685)", {
+  l = lrn("classif.debug", save_tasks = TRUE)$train(tsk("iris")$select(c("Sepal.Length", "Sepal.Width")))
+  expect_error(l$predict_newdata(data.table(Sepal.Length = 1, Sepal.Width = "abc")),
+    "Incompatible types during auto-converting column 'Sepal.Width'", fixed = TRUE)
+  expect_error(l$predict_newdata(data.table(Sepal.Length = 1L)),
+    "but is missing elements")
+
+  # New test for integerish value conversion to double
+  p1 = l$predict_newdata(data.table(Sepal.Length = 1, Sepal.Width = 2))
+  p2 = l$predict_newdata(data.table(Sepal.Length = 1L, Sepal.Width = 2))
+  expect_equal(l$model$task_predict$col_info[list("Sepal.Length")]$type, "numeric")
+  expect_double(l$model$task_predict$data(cols = "Sepal.Length")[[1]])
+
+  expect_equal(p1, p2)
+})