[R-package] Turn matrix to storage mode "double" (microsoft#3140)

* Turn matrix to storage mode "double"

* added "test_Predictor.R" to R tests with check for integer storage mode during prediction

* Explicit integers in test_Preditor to avoid TravisCI failure

* properly aligning comment on storage.mode

Co-authored-by: James Lamb <jaylamb20@gmail.com>

* split double assignment into two lines

Co-authored-by: James Lamb <jaylamb20@gmail.com>

R-package/R/lgb.Predictor.R

@@ -142,6 +142,10 @@ Predictor <- R6::R6Class(
 
         # Check if data is a matrix
         if (is.matrix(data)) {
+          # Check whether matrix is the correct type first ("double")
+          if (storage.mode(data) != "double") {
+            storage.mode(data) <- "double"
+          }
           preds <- lgb.call(
             "LGBM_BoosterPredictForMat_R"
             , ret = preds

R-package/tests/testthat/test_Predictor.R

@@ -0,0 +1,18 @@
+context("Predictor")
+
+test_that("predictions do not fail for integer input", {
+    X <- as.matrix(as.integer(iris[, "Species"]), ncol = 1L)
+    y <- iris[["Sepal.Length"]]
+    dtrain <- lgb.Dataset(X, label = y)
+    fit <- lgb.train(
+        data = dtrain
+        , objective = "regression"
+        , verbose = -1L
+    )
+    X_double <- X[c(1L, 51L, 101L), , drop = FALSE]
+    X_integer <- X_double
+    storage.mode(X_double) <- "double"
+    pred_integer <- predict(fit, X_integer)
+    pred_double <- predict(fit, X_double)
+    expect_equal(pred_integer, pred_double)
+})