'geom_rug()' prints a warning when 'na.rm = FALSE' (#5906)

* 'geom_rug()' prints a warning when 'na.rm = FALSE'

Fixes issue #5905.
When presented with missing values, 'geom_rug()' was not printing a
warning message, contrary to the documentation.
A warning message is now printed when 'na.rm = FALSE' ad suppressed when
'na.rm = TRUE', as expected.

* Impement local changes in 'handle_na()'

Make local changes in 'handle_na() instead of global changes to 'GeomRug'

* Test for 'geom_rug()' warning about missing values

Also added the issue number to NEWS

* Handle rugs in orthogonal directions better

When plotting rugs in both the 'x' and 'y' direction simultaneously,
values of 'x' were being dropped when 'y' was missing, and vice versa.
A warning will be given for each axis ('x' or 'y') that contains missing
values, if 'na.rm = FALSE'.

* Remove dependence on 'dplyr'

Use 'vctrs::vec_set_union()' instead of 'dplyr::union()'

* Update 'testthat' version to 3.1.5

* Add github handle to NEWS bullet

Author: pn317

DESCRIPTION

@@ -67,7 +67,7 @@ Suggests:
     rpart,
     sf (>= 0.7-3),
     svglite (>= 2.1.2),
-    testthat (>= 3.1.2),
+    testthat (>= 3.1.5),
     vdiffr (>= 1.0.6),
     xml2
 Enhances: 

NEWS.md

@@ -1,5 +1,6 @@
 # ggplot2 (development version)
 
+* `geom_rug()` prints a warning when `na.rm = FALSE`, as per documentation (@pn317, #5905)
 * `position_dodge(preserve = "single")` now handles multi-row geoms better,
   such as `geom_violin()` (@teunbrand based on @clauswilke's work, #2801).
 * `position_jitterdodge()` now dodges by `group` (@teunbrand, #3656)

R/geom-rug.R

@@ -157,5 +157,44 @@ GeomRug <- ggproto("GeomRug", Geom,
 
   draw_key = draw_key_path,
 
-  rename_size = TRUE
+  rename_size = TRUE,
+
+  handle_na = function(self, data, params) {
+    sides_aes <- character()
+
+    if (grepl("b|t", params$sides)) {
+      sides_aes <- c(sides_aes, "x")
+    }
+
+    if (grepl("l|r", params$sides)) {
+      sides_aes <- c(sides_aes, "y")
+    }
+
+    if (length(sides_aes) > 0) {
+      df_list <- lapply(
+        sides_aes,
+        function(axis) {
+          remove_missing(
+            data, params$na.rm,
+            c(axis, self$required_aes, self$non_missing_aes),
+            snake_class(self)
+          )
+        }
+      )
+      data <- switch(
+        paste0(sides_aes, collapse = ""),
+        "x" = ,
+        "y" = df_list[[1]],
+        "xy" = vctrs::vec_set_union(df_list[[1]], df_list[[2]])
+      )
+    } else {
+      data <- remove_missing(
+        data, params$na.rm,
+        c(self$required_aes, self$non_missing_aes),
+        snake_class(self)
+      )
+    }
+
+    data
+  }
 )

tests/testthat/test-geom-rug.R

@@ -40,3 +40,21 @@ test_that("Rug lengths are correct", {
 
 })
 
+test_that(
+  "geom_rug() warns about missing values when na.rm = FALSE",
+  {
+    df2 <- df
+    n_missing <- 2
+    df2$x[sample(nrow(df2), size = n_missing)] <- NA
+
+    p1 <- ggplot(df2, aes(x = x)) + geom_rug()
+    p2 <- ggplot(df2, aes(x = x)) + geom_rug(na.rm = TRUE)
+
+    expect_warning(
+      ggplotGrob(p1),
+      paste0("Removed ", n_missing, " rows containing missing values or values outside the scale range")
+    )
+
+    expect_no_warning(ggplotGrob(p2))
+  }
+)