Scales don't squish to zero-range domains (#5279)

* Don't squish to zero-range domains

* Add test battery

Author: teunbrand

R/scale-.R

@@ -648,7 +648,7 @@ ScaleContinuous <- ggproto("ScaleContinuous", Scale,
     domain <- suppressWarnings(self$trans$transform(self$trans$domain))
     domain <- sort(domain)
     # To avoid NaN causing issues. NaN are dropped by the sort()
-    if (length(domain) == 2) {
+    if (length(domain) == 2 && !zero_range(domain)) {
       limits <- oob_squish(limits, domain)
     }
 

tests/testthat/_snaps/scales.md

@@ -42,3 +42,10 @@
 
     `breaks` and `labels` are different lengths
 
+# numeric scale transforms can produce breaks
+
+    Code
+      test_breaks("log", limits = c(0.1, 1000))
+    Output
+      [1]        NA   1.00000  20.08554 403.42879
+

tests/testthat/test-scales.R

@@ -468,3 +468,67 @@ test_that("staged aesthetics are backtransformed properly (#4155)", {
   # x / 2 should be 16 / 2 = 8, thus the result should be sqrt(8) on scale_x_sqrt()
   expect_equal(layer_data(p)$x, sqrt(8))
 })
+
+test_that("numeric scale transforms can produce breaks", {
+
+  test_breaks <- function(trans, limits) {
+    scale <- scale_x_continuous(trans = trans)
+    scale$train(scale$transform(limits))
+    view <- view_scale_primary(scale)
+    scale$trans$inverse(view$get_breaks())
+  }
+
+  expect_equal(test_breaks("asn", limits = c(0, 1)),
+               seq(0, 1, by = 0.25))
+
+  expect_equal(test_breaks("sqrt", limits = c(0, 10)),
+               seq(0, 10, by = 2.5))
+
+  expect_equal(test_breaks("atanh", limits = c(-0.9, 0.9)),
+               c(NA, -0.5, 0, 0.5, NA))
+
+  # Broken, should fix on {scale}'s side
+  # expect_equal(test_breaks(boxcox_trans(0), limits = c(0, 10)), ...)
+
+  expect_equal(test_breaks(modulus_trans(0), c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  expect_equal(test_breaks(yj_trans(0), c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  expect_equal(test_breaks("exp", c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  expect_equal(test_breaks("identity", limits = c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  # irrational numbers, so snapshot values
+  expect_snapshot(test_breaks("log", limits = c(0.1, 1000)))
+
+  expect_equal(test_breaks("log10", limits = c(0.1, 1000)),
+               10 ^ seq(-1, 3))
+
+  expect_equal(test_breaks("log2", limits = c(0.5, 32)),
+               c(0.5, 2, 8, 32))
+
+  expect_equal(test_breaks("log1p", limits = c(0, 10)),
+               seq(0, 10, by = 2.5))
+
+  expect_equal(test_breaks("pseudo_log", limits = c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  expect_equal(test_breaks("logit", limits = c(0.001, 0.999)),
+               c(NA, 0.25, 0.5, 0.75, NA))
+
+  expect_equal(test_breaks("probit", limits = c(0.001, 0.999)),
+               c(NA, 0.25, 0.5, 0.75, NA))
+
+  expect_equal(test_breaks("reciprocal", limits = c(1, 10)),
+               c(NA, 2.5, 5, 7.5, 10))
+
+  expect_equal(test_breaks("reverse", limits = c(-10, 10)),
+               seq(-10, 10, by = 5))
+
+  expect_equal(test_breaks("sqrt", limits = c(0, 10)),
+               seq(0, 10, by = 2.5))
+})