Merge pull request #5139 from teunbrand/ordered_vec_rbind0

Try coerce ordered in `vec_rbind0()`

Author: teunbrand

NEWS.md

@@ -1,5 +1,8 @@
 # ggplot2 (development version)
 
+* Using two ordered factors as facetting variables in 
+  `facet_grid(..., as.table = FALSE)` now throws a warning instead of an
+  error (@teunbrand, #5109).
 * Added `scale_linewidth_manual()` and `scale_linewidth_identity()` to support
   the `linewidth` aesthetic (@teunbrand, #5050).
 * Automatic breaks in `scale_*_binned()` should no longer be out-of-bounds,

R/utilities.r

@@ -633,7 +633,8 @@ with_ordered_restart <- function(expr, .call) {
         restart <- TRUE
         if (is.ordered(x)) {
           x <- factor(as.character(x), levels = levels(x))
-        } else {
+        }
+        if (is.ordered(y)) {
           y <- factor(as.character(y), levels = levels(y))
         }
       } else if (is.character(x) || is.character(y)) {

tests/testthat/test-utilities.r

@@ -127,3 +127,34 @@ test_that("cut_*() checks its input and output", {
 test_that("interleave() checks the vector lengths", {
   expect_snapshot_error(interleave(1:4, numeric()))
 })
+
+test_that("vec_rbind0 can combined ordered factors", {
+
+  withr::local_options(lifecycle_verbosity = "warning")
+
+  # Ideally code below throws just 1 warning (the <ordered> and <ordered> one)
+  # However, it was technically challenging to reduce the numbers of warnings
+  # See #5139 for more details
+
+  expect_warning(
+    expect_warning(
+      expect_warning(
+        {
+          test <- vec_rbind0(
+            data_frame0(a = factor(c("A", "B"), ordered = TRUE)),
+            data_frame0(a = factor(c("B", "C"), ordered = TRUE))
+          )
+        },
+        "<ordered> and <ordered>", class = "lifecycle_warning_deprecated"
+      ),
+      "<ordered> and <factor>", class = "lifecycle_warning_deprecated"
+    ),
+    "<ordered> and <factor>", class = "lifecycle_warning_deprecated"
+  )
+
+  # Should be <factor> not <ordered/factor>, hence the 'exact'
+  expect_s3_class(test$a, "factor", exact = TRUE)
+  # Test levels are combined sensibly
+  expect_equal(levels(test$a), c("A", "B", "C"))
+
+})