use correct version for xgboost switching

Author: EmilHvitfeldt

R/boost_tree.R

@@ -340,7 +340,7 @@ xgb_train <- function(
 
   others <- process_others(others, arg_list)
 
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     if (!is.null(num_class) && num_class > 2) {
       arg_list$num_class <- num_class
     }
@@ -380,7 +380,7 @@ xgb_train <- function(
     ),
     others
   )
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     main_args$evals <- quote(x$watchlist)
   } else {
     main_args$watchlist <- quote(x$watchlist)
@@ -543,7 +543,7 @@ as_xgb_data <- function(
       watch_list <- list(validation = val_data)
 
       info_list <- list(label = y[trn_index])
-      if (utils::packageVersion("xgboost") > "2.0.0.0") {
+      if (utils::packageVersion("xgboost") >= "2.0.0.0") {
         if (!is.null(weights)) {
           dat <- xgboost::xgb.DMatrix(
             data = x[trn_index, , drop = FALSE],
@@ -569,7 +569,7 @@ as_xgb_data <- function(
         )
       }
     } else {
-      if (utils::packageVersion("xgboost") > "2.0.0.0") {
+      if (utils::packageVersion("xgboost") >= "2.0.0.0") {
         if (!is.null(weights)) {
           dat <- xgboost::xgb.DMatrix(
             x,
@@ -647,7 +647,7 @@ multi_predict._xgb.Booster <-
   }
 
 xgb_by_tree <- function(tree, object, new_data, type, ...) {
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     pred <- xgb_predict(
       object$fit,
       new_data = new_data,

tests/testthat/test-boost_tree_xgboost.R

@@ -9,7 +9,7 @@ hpc_xgboost <-
   set_engine("xgboost")
 
 extract_xgb_param <- function(x, param) {
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     res <- attr(extract_fit_engine(x), "params")[[param]]
   } else {
     res <- extract_fit_engine(x)$param[[param]]
@@ -18,7 +18,7 @@ extract_xgb_param <- function(x, param) {
 }
 
 extract_xgb_evaluation_log <- function(x) {
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     res <- attr(extract_fit_engine(x), "evaluation_log")
   } else {
     res <- extract_fit_engine(x)[["evaluation_log"]]
@@ -408,15 +408,15 @@ test_that('early stopping', {
   )
 
   extract_xgb_nitter <- function(x) {
-    if (utils::packageVersion("xgboost") > "2.0.0.0") {
+    if (utils::packageVersion("xgboost") >= "2.0.0.0") {
       res <- nrow(attr(extract_fit_engine(x), "evaluation_log"))
     } else {
       res <- extract_fit_engine(reg_fit)$niter
     }
     res
   }
   extract_xgb_best_iteration <- function(x) {
-    if (utils::packageVersion("xgboost") > "2.0.0.0") {
+    if (utils::packageVersion("xgboost") >= "2.0.0.0") {
       res <- attr(extract_fit_engine(x), "early_stop")$best_iteration
     } else {
       res <- extract_fit_engine(reg_fit)$best_iteration
@@ -572,7 +572,7 @@ test_that('xgboost data and sparse matrices', {
   from_mat$fit$handle <- NULL
   from_sparse$fit$handle <- NULL
 
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     expect_equal(
       attributes(extract_fit_engine(from_df)),
       attributes(extract_fit_engine(from_mat)),
@@ -689,7 +689,7 @@ test_that("fit and prediction with `event_level`", {
   xgbmat_train_1 <- xgb.DMatrix(data = train_x, label = train_y_1)
 
   set.seed(24)
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     fit_xgb_1 <- xgboost::xgb.train(
       params = list(
         objective = "binary:logistic",
@@ -729,7 +729,7 @@ test_that("fit and prediction with `event_level`", {
   xgbmat_train_2 <- xgb.DMatrix(data = train_x, label = train_y_2)
 
   set.seed(24)
-  if (utils::packageVersion("xgboost") > "2.0.0.0") {
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
     fit_xgb_2 <- xgboost::xgb.train(
       params = list(
         eval_metric = "auc",