Make work with newer versions of xgboost

NEWS.md

@@ -4,6 +4,7 @@
 
 * `surv_reg()` is now defunct and will error if called. Please use `survival_reg()` instead (#1206).
 
+* Enable parsnip to work with xgboost version > 2.0.0.0. (#1227)
 
 # parsnip 1.3.3
 

R/boost_tree.R

@@ -271,6 +271,7 @@ xgb_train <- function(
   event_level = c("first", "second"),
   ...
 ) {
+  rlang::check_installed("xgboost")
   event_level <- rlang::arg_match(event_level, c("first", "second"))
   others <- list(...)
 
@@ -340,31 +341,70 @@ xgb_train <- function(
 
   others <- process_others(others, arg_list)
 
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
+    if (!is.null(num_class) && num_class > 2) {
+      arg_list$num_class <- num_class
+    }
+
+    param_names <- names(
+      formals(
+        getFromNamespace("xgb.params", ns = "xgboost")
+      )
+    )
+
+    if (any(param_names %in% names(others))) {
+      elements <- param_names[param_names %in% names(others)]
+
+      for (element in elements) {
+        arg_list[[element]] <- others[[element]]
+        others[[element]] <- NULL
+      }
+    }
+
+    if (is.null(arg_list$objective)) {
+      if (is.numeric(y)) {
+        arg_list$objective <- "reg:squarederror"
+      } else {
+        if (num_class == 2) {
+          arg_list$objective <- "binary:logistic"
+        } else {
+          arg_list$objective <- "multi:softprob"
+        }
+      }
+    }
+  }
+
   main_args <- c(
     list(
       data = quote(x$data),
-      watchlist = quote(x$watchlist),
       params = arg_list,
       nrounds = nrounds,
       early_stopping_rounds = early_stop
     ),
     others
   )
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
+    main_args$evals <- quote(x$watchlist)
+  } else {
+    main_args$watchlist <- quote(x$watchlist)
+  }
 
-  if (is.null(main_args$objective)) {
-    if (is.numeric(y)) {
-      main_args$objective <- "reg:squarederror"
-    } else {
-      if (num_class == 2) {
-        main_args$objective <- "binary:logistic"
+  if (utils::packageVersion("xgboost") < "2.0.0.0") {
+    if (is.null(main_args$objective)) {
+      if (is.numeric(y)) {
+        main_args$objective <- "reg:squarederror"
       } else {
-        main_args$objective <- "multi:softprob"
+        if (num_class == 2) {
+          main_args$objective <- "binary:logistic"
+        } else {
+          main_args$objective <- "multi:softprob"
+        }
       }
     }
-  }
 
-  if (!is.null(num_class) && num_class > 2) {
-    main_args$num_class <- num_class
+    if (!is.null(num_class) && num_class > 2) {
+      main_args$num_class <- num_class
+    }
   }
 
   call <- make_call(fun = "xgb.train", ns = "xgboost", main_args)
@@ -471,6 +511,7 @@ as_xgb_data <- function(
   event_level = "first",
   ...
 ) {
+  rlang::check_installed("xgboost")
   lvls <- levels(y)
   n <- nrow(x)
 
@@ -506,21 +547,52 @@ as_xgb_data <- function(
       watch_list <- list(validation = val_data)
 
       info_list <- list(label = y[trn_index])
-      if (!is.null(weights)) {
-        info_list$weight <- weights[trn_index]
+      if (utils::packageVersion("xgboost") >= "2.0.0.0") {
+        if (!is.null(weights)) {
+          dat <- xgboost::xgb.DMatrix(
+            data = x[trn_index, , drop = FALSE],
+            missing = NA,
+            label = y[trn_index],
+            weight = weights[trn_index]
+          )
+        } else {
+          dat <- xgboost::xgb.DMatrix(
+            data = x[trn_index, , drop = FALSE],
+            missing = NA,
+            label = y[trn_index]
+          )
+        }
+      } else {
+        if (!is.null(weights)) {
+          info_list$weight <- weights[trn_index]
+        }
+        dat <- xgboost::xgb.DMatrix(
+          data = x[trn_index, , drop = FALSE],
+          missing = NA,
+          info = info_list
+        )
       }
-      dat <- xgboost::xgb.DMatrix(
-        data = x[trn_index, , drop = FALSE],
-        missing = NA,
-        info = info_list
-      )
     } else {
-      info_list <- list(label = y)
-      if (!is.null(weights)) {
-        info_list$weight <- weights
+      if (utils::packageVersion("xgboost") >= "2.0.0.0") {
+        if (!is.null(weights)) {
+          dat <- xgboost::xgb.DMatrix(
+            x,
+            missing = NA,
+            label = y,
+            weight = weights
+          )
+        } else {
+          dat <- xgboost::xgb.DMatrix(x, missing = NA, label = y)
+        }
+        watch_list <- list(training = dat)
+      } else {
+        info_list <- list(label = y)
+        if (!is.null(weights)) {
+          info_list$weight <- weights
+        }
+        dat <- xgboost::xgb.DMatrix(x, missing = NA, info = info_list)
+        watch_list <- list(training = dat)
       }
-      dat <- xgboost::xgb.DMatrix(x, missing = NA, info = info_list)
-      watch_list <- list(training = dat)
     }
   } else {
     dat <- xgboost::setinfo(x, "label", y)
@@ -579,12 +651,21 @@ multi_predict._xgb.Booster <-
   }
 
 xgb_by_tree <- function(tree, object, new_data, type, ...) {
-  pred <- xgb_predict(
-    object$fit,
-    new_data = new_data,
-    iterationrange = c(1, tree + 1),
-    ntreelimit = NULL
-  )
+  rlang::check_installed("xgboost")
+  if (utils::packageVersion("xgboost") >= "2.0.0.0") {
+    pred <- xgb_predict(
+      object$fit,
+      new_data = new_data,
+      iterationrange = c(1, tree + 1)
+    )
+  } else {
+    pred <- xgb_predict(
+      object$fit,
+      new_data = new_data,
+      iterationrange = c(1, tree + 1),
+      ntreelimit = NULL
+    )
+  }
 
   # switch based on prediction type
   if (object$spec$mode == "regression") {

man/rmd/boost_tree_xgboost.Rmd

@@ -3,7 +3,7 @@
 #| include: false
 ```
 
-`r descr_models("boost_tree", "xgboost")`
+`r descr_models("boost_tree", "xgboost")`. Note that in late 2025, a new version of xgboost was released with differences in its interface and model objects. This version of parsnip should work with either version. 
 
 ## Tuning Parameters
 

man/rmd/boost_tree_xgboost.md

@@ -1,7 +1,7 @@
 
 
 
-For this engine, there are multiple modes: classification and regression
+For this engine, there are multiple modes: classification and regression. Note that in late 2025, a new version of xgboost was released with differences in its interface and model objects. This version of parsnip should work with either version. 
 
 ## Tuning Parameters