geom_raster() fallback for non-Cartesian coords

NEWS.md

@@ -1,5 +1,7 @@
 # ggplot2 (development version)
 
+* `geom_raster()` now falls back to rendering as `geom_rect()` when coordinates
+  are not Cartesian (#5503).
 * `stat_ecdf()` now has an optional `weight` aesthetic (@teunbrand, #5058).
 * Position scales combined with `coord_sf()` can now use functions in the 
  `breaks` argument. In addition, `n.breaks` works as intended and 

R/geom-raster.R

@@ -88,9 +88,14 @@ GeomRaster <- ggproto("GeomRaster", Geom,
   draw_panel = function(self, data, panel_params, coord, interpolate = FALSE,
                         hjust = 0.5, vjust = 0.5) {
     if (!inherits(coord, "CoordCartesian")) {
-      cli::cli_abort(c(
-        "{.fn {snake_class(self)}} only works with {.fn coord_cartesian}."
+      cli::cli_inform(c(
+        "{.fn {snake_class(self)}} only works with {.fn coord_cartesian}.",
+        i = "Falling back to drawing as {.fn {snake_class(GeomRect)}}."
       ))
+      data$linewidth <- 0.3 # preventing anti-aliasing artefacts
+      data$colour <- data$fill
+      grob <- GeomRect$draw_panel(data, panel_params, coord)
+      return(grob)
     }
 
     # Convert vector of data to raster

tests/testthat/_snaps/geom-raster.md

@@ -14,13 +14,6 @@
 
     `vjust` must be a number, not the string "a".
 
----
-
-    Problem while converting geom to grob.
-    i Error occurred in the 1st layer.
-    Caused by error in `draw_panel()`:
-    ! `geom_raster()` only works with `coord_cartesian()`.
-
 # geom_raster() fails with pattern fills
 
     Problem while converting geom to grob.

tests/testthat/test-geom-raster.R

@@ -6,7 +6,7 @@ test_that("geom_raster() checks input and coordinate system", {
 
   df <- data_frame(x = rep(c(-1, 1), each = 3), y = rep(-1:1, 2), z = 1:6)
   p <- ggplot(df, aes(x, y, fill = z)) + geom_raster() + coord_polar()
-  expect_snapshot_error(ggplotGrob(p))
+  expect_message(ggplotGrob(p), "only works with")
 })
 
 test_that("geom_raster() fails with pattern fills", {
@@ -66,6 +66,14 @@ test_that("geom_raster draws correctly", {
       geom_point(colour = "red")
   )
 
+  # In non-linear coordinates
+  df <- data.frame(x = c(1, 2, 1, 2), y = c(1, 1, 2, 2), fill = LETTERS[1:4])
+  suppressMessages(
+    expect_doppelganger("rectangle fallback",
+      ggplot(df, aes(x, y, fill = fill)) + geom_raster() + coord_polar()
+    )
+  )
+
   # Categorical fill, irregular swatches ---------------------------------------
 
   df <- expand.grid(x = 1:10, y = 1:10)