Surprising failure with xlim

[David Hugh-Jones]

Example:

library(ggplot2)
ggp <- ggplot(iris, aes(Sepal.Width, Sepal.Length, color = Species, shape = Species))
ggp2 <- ggp + geom_point()

from <- c(2.5, 3.0, 5.5, 6.0)
to <- c(3.5, 4.5, 4.5, 5.5)
ggp2 + xlim(2,5) + geom_magnify(aes(from = Species=="versicolor" & Sepal.Length < 6), to = c(4, 4.8, 5, 6))

This doesn't print a magnified inset, and gives a warning about dropped cases.

The failure is because xmin becomes NA in the data, which then leads all cases to be dropped as incomplete.

Not sure why xmin is becoming NA. It's fine with xlim(1.9, 5) ; or with expand = 0.

[David Hugh-Jones]

StatMagnify unsurprisingly sets xmin to 1.94. Then some time before the geom, that 1.94 becomes NA in the data.