Fix empty dataframe error with process_md = TRUE

R/dt_stub_df.R

@@ -145,7 +145,12 @@ dt_stub_df_init <- function(
 
       # dplyr::recode is superseded, and is slower now.
       # TODO consider using vctrs::vec_case_match when available r-lib/vctrs#1622
-      row_group_ids <- dplyr::recode(row_group_labels, !!!unique_row_group_ids)
+      # Handle empty dataframe case where unique_row_group_ids is empty
+      if (length(unique_row_group_ids) == 0) {
+        row_group_ids <- row_group_labels
+      } else {
+        row_group_ids <- dplyr::recode(row_group_labels, !!!unique_row_group_ids)
+      }
 
     } else {
       row_group_ids <- row_group_labels

tests/testthat/test-gt_object.R

@@ -992,3 +992,34 @@ test_that("Formatting functions operate with a 'last-one-wins' approach", {
       )
     )
 })
+
+test_that("Empty dataframes with process_md = TRUE work correctly (issue #2081)", {
+
+  # Test case 1: Empty dataframe with groupname_col and process_md = TRUE
+  # Should not throw "No replacements provided" error
+  expect_no_error(
+    mtcars[0, ] |> gt(groupname_col = "cyl", process_md = TRUE)
+  )
+
+  # Test case 2: Empty grouped dataframe with process_md = TRUE
+  # Should not throw "No replacements provided" error
+  expect_no_error(
+    mtcars[0, ] |> dplyr::group_by(cyl) |> gt(process_md = TRUE)
+  )
+
+  # Test case 3: Empty dataframe with rowname_col and process_md = TRUE
+  # Should continue to work as before
+  expect_no_error(
+    mtcars[0, ] |> gt(rowname_col = "cyl", process_md = TRUE)
+  )
+
+  # Test case 4: Regression test - non-empty dataframes should still work
+  expect_no_error(
+    mtcars[1:5, ] |> gt(groupname_col = "cyl", process_md = TRUE)
+  )
+
+  # Test case 5: Verify empty table is created with correct structure
+  empty_gt <- mtcars[0, ] |> gt(groupname_col = "cyl", process_md = TRUE)
+  expect_s3_class(empty_gt, "gt_tbl")
+  expect_equal(nrow(dt_data_get(empty_gt)), 0)
+})