diff --git a/R/pkg/NAMESPACE b/R/pkg/NAMESPACE
index 82e56ca43729..1bc27ba88fe0 100644
--- a/R/pkg/NAMESPACE
+++ b/R/pkg/NAMESPACE
@@ -258,6 +258,7 @@ exportMethods("%in%",
               "skewness",
               "sort_array",
               "soundex",
+              "spark_partition_id",
               "stddev",
               "stddev_pop",
               "stddev_samp",
diff --git a/R/pkg/R/functions.R b/R/pkg/R/functions.R
index a779127b379a..064edf2468dc 100644
--- a/R/pkg/R/functions.R
+++ b/R/pkg/R/functions.R
@@ -1179,6 +1179,27 @@ setMethod("soundex",
             column(jc)
           })
 
+#' Return the partition ID as a column
+#'
+#' Return the partition ID of the Spark task as a SparkDataFrame column.
+#' Note that this is nondeterministic because it depends on data partitioning and
+#' task scheduling.
+#'
+#' This is equivalent to the SPARK_PARTITION_ID function in SQL.
+#'
+#' @rdname spark_partition_id
+#' @name spark_partition_id
+#' @export
+#' @examples
+#' \dontrun{select(df, spark_partition_id())}
+#' @note spark_partition_id since 2.0.0
+setMethod("spark_partition_id",
+          signature(x = "missing"),
+          function() {
+            jc <- callJStatic("org.apache.spark.sql.functions", "spark_partition_id")
+            column(jc)
+          })
+
 #' @rdname sd
 #' @name stddev
 setMethod("stddev",
diff --git a/R/pkg/R/generics.R b/R/pkg/R/generics.R
index 6e754afab6c6..8abd18d35e66 100644
--- a/R/pkg/R/generics.R
+++ b/R/pkg/R/generics.R
@@ -1126,6 +1126,10 @@ setGeneric("sort_array", function(x, asc = TRUE) { standardGeneric("sort_array")
 #' @export
 setGeneric("soundex", function(x) { standardGeneric("soundex") })
 
+#' @rdname spark_partition_id
+#' @export
+setGeneric("spark_partition_id", function(x) { standardGeneric("spark_partition_id") })
+
 #' @rdname sd
 #' @export
 setGeneric("stddev", function(x) { standardGeneric("stddev") })
diff --git a/R/pkg/inst/tests/testthat/test_sparkSQL.R b/R/pkg/inst/tests/testthat/test_sparkSQL.R
index fcc2ab3ed6a2..53097f34d7a3 100644
--- a/R/pkg/inst/tests/testthat/test_sparkSQL.R
+++ b/R/pkg/inst/tests/testthat/test_sparkSQL.R
@@ -1058,6 +1058,7 @@ test_that("column functions", {
   c16 <- is.nan(c) + isnan(c) + isNaN(c)
   c17 <- cov(c, c1) + cov("c", "c1") + covar_samp(c, c1) + covar_samp("c", "c1")
   c18 <- covar_pop(c, c1) + covar_pop("c", "c1")
+  c19 <- spark_partition_id()
 
   # Test if base::is.nan() is exposed
   expect_equal(is.nan(c("a", "b")), c(FALSE, FALSE))