Have variable distance between geom_boxplot/geom_violine based on x-axis values

Question

Problem

I have an x variable - for instance size that is meaningful so that something can be two times the size of something else. I want to relate x to y (some other variable). At the same time, due to sampling x does not vary continuously but is discrete because there are just a few different object types and all objects of the same type of the same size (e.g. size of 1, 3 or 10). I want to use geom's like geom_boxplot or geom_violin to dsiplay the relationship between x & y.

However, the problem is that: If I keep x numeric, then I am only getting one boxplot/violin. If I convert it to a factor (shown below), then the distance between the geom does not reflect the distance in x. For instance, the distance between 1 & 3 is the same as the distance between 3 & 10.

Is there a way to discretise the data but change the spacing so it reflects the actual difference on the x-axis and use those geoms?

Code

# Seed for reproducibility
set.seed(20230518)

# Create random data
n  <- 10
df <- data.frame(x = factor(rep(c(1, 3, 10), each = n)),
                 y = c(rnorm(n), rnorm(n), rnorm(n)))


# Box plot version
ggplot(df, aes(x = x, y = y)) + geom_boxplot() + geom_point()

# Violine plot verion
ggplot(df, aes(x = x, y = y)) + geom_violin() + geom_point()

Boxplots

Violins

Expected solution

The distance between the geoms reflects the difference in x. This be should similar to this just with geom_boxplot & geom_violin in addition to the points:

# Nnumeric
ggplot(df, aes(x = as.numeric(as.character(x)), y = y)) + geom_point()

Answer 1

Add missing factor levels, then set drop to FALSE, works for geom_violin, too.

df$x <- factor(df$x, levels = 1:10)

ggplot(df, aes(x = x, y = y)) + 
  geom_boxplot() + 
  geom_point() +
  scale_x_discrete(drop = FALSE)

Add breaks to hide other x values:

scale_x_discrete(drop = FALSE, breaks = unique(sort(df$x)))