Quirky NA-handling of `geom_ribbon()`

[teunbrand]

Now that we're allowing gradients in geom_ribbon() and derivatives, the following caught my attention.

In the example below, we simply want to vary the gradient along x, so top and bottom gradients should match.
However, the gradient is misplaced because the second group has a missing value (note it starts with dark blue in the middle).

devtools::load_all("~/packages/ggplot2/")
#> ℹ Loading ggplot2

df <- data.frame(
  x = c(1:3, 1:3),
  ymin = c(1, 1, 1, NA, 2, 2),
  ymax = c(1.5, 2, 1.5, 2.5, 3, 2.5),
  group = c(1,1,1,2,2,2)
)

p <- ggplot(df, aes(x, ymin = ymin, ymax = ymax, group = group, fill = x)) +
  scale_fill_viridis_c()

# Wrong
p + geom_ribbon()

The gradients are drawn correctly if missing values are removed properly.

p + geom_ribbon(na.rm = TRUE)

^{Created on 2024-12-16 with reprex v2.1.1}

I think there are two things wrong with this when na.rm = FALSE:

• There is no warning about missing values, as you would expect with any other geom
• It doesn't actually remove the missing data, causing the misalignment of gradients.

I traced this decision to a remedy against #1549.

[clauswilke]

In either case, this seems to be an incorrect implementation of na.rm. In all other geoms it has no effect other than suppressing a warning. But here it seems to alter the data.

From the geom_point() documentation (as one example):

[teunbrand]

Agreed. If #1549 it intended to not fill over missing values, we could separate groups of NA-gaps.