如何才能使我的缩放转换点不会被ggplot2剪切？

Question

I'm making a dotchart where I want to log transform the x scale. I have some points at 0, which ggplot handles nicely in that it doesn't eliminate them, however it does clip them.

What can I do so the points at 0 don't get clipped? It seems xlim() and the scale transformation don't play together, only the last one called takes effect.

An example:

myData <- data.frame(x = c(rexp(5), 0), y = "category")
myBreaks <- c(.1, 1, 5)
ggplot(myData, aes(x = x, y = y)) +
scale_x_continuous(trans = "log",
                   breaks = myBreaks,
                   labels = myBreaks) +
geom_point(size = 5, legend = F)

Answer 1

Since log(0) is -Inf , I suspect that your 0 point will always be clipped if you keep it zero. I tried to fiddle with expand=... , coord_trans and everything else I could think of.

Here is a workaround:

• set your zero values to an arbitrary small value (say 1e-6)
• include a break at that value
• optionally label that break 0

The code:

myData <- data.frame(x = c(rexp(5), 0), y = "category")

myData <- within(myData, x[x==0] <- 1e-6)
myBreaks <- c(1e-6, 0.1, 1, 5)
myLabels <- c(0, myBreaks[-1])
ggplot(myData, aes(x = x, y = y)) +
    geom_point(size = 5, legend = F) +
    scale_x_continuous(
        trans = "log",
        breaks = myBreaks,
        labels = myLabels
    ) 

Answer 2

With the release of ggplot2 version 3.0.0, you can use coord_cartesian(clip = 'off') .

library(ggplot2)

myData <- data.frame(x = c(rexp(5), 0), y = "category")
myBreaks <- c(.1, 1, 5)
ggplot(myData, aes(x = x, y = y)) +
  scale_x_continuous(trans = "log",
                     breaks = myBreaks,
                     labels = myBreaks) +
  geom_point(size = 5, legend = F) +
  coord_cartesian(clip = 'off') +
  labs(title = "coord_cartesian(clip = 'off')")