geom_jitter fails when coordinate values have very small spread

[nicholasdavies]

When geom_jitter is given some data where the y coordinates are all the same, a default jitter height is chosen. But if the y coordinates differ only by a small magnitude, the selection of the jitter height seems to fail and nothing is plotted.

Minimal working example:

library(ggplot2)

set.seed(123)
ggplot(data.frame(x = 1:50, y = 0)) +
    geom_jitter(aes(x, y))

set.seed(123)
ggplot(data.frame(x = 1:50, y = 1e-8 * rnorm(50))) +
    geom_jitter(aes(x, y))
#> Warning in min(d[d > tolerance]): no non-missing arguments to min; returning
#> Inf
#> Warning in stats::runif(length(x), -amount, amount): NAs produced
#> Warning: Removed 50 rows containing missing values or values outside the scale range
#> (`geom_point()`).

^{Created on 2025-07-25 with reprex v2.1.1}

I would have expected both plots to be more or less the same.

[teunbrand]

Thanks for the report! I can reproduce the issue.
I'd believe the solution to this issue is building a contingency for when none of the values exceed the tolerance here:

ggplot2/R/utilities-resolution.R

Lines 35 to 36 in bed8e32

	tolerance <- sqrt(.Machine$double.eps)
	min(d[d > tolerance])