ggplot: rearrange factor order and highlight area

Question

First, the way to specify the order of categorical variables for ggplot is to reorder its levels in the data.frame. Second, to have an area highlighted on any plot we can use geom_rect. Here we need to note that it is a key thing not to pass the data to geom_rect, otherwise it does not let us to set alpha, thus the gridlines won't be visible. Now there are 2 cases:

• if we pass the data to geom_rect (or to the top level ggplot()), the order agrees with that in the data.frame, but as I mentioned the rectangle won't be transparent
• if we pass the data only to the geom_point layer, ggplot rearranges the discrete variables in alphabetical order

How to have both of the two criteria, i.e. have the predefined order and have a transparent rectangle in the desired position?

Bonus question: how to have a rectangle at discrete variables with its edges between gridlines, i.e. adjusted by 0.5? vjust and hjust are not used arguments here (as a warning tells us). And how to make rectangle filling the whole vertical space (for this we would need to define ymax as the n+1th factor level, which does not exist).

require(ggplot2)

ex <- data.frame(a = factor(letters[1:10]),
                 b = factor(rep(c('b', 'a'), 5)),
                 c = rep(letters[1:5], 2))

ex$a <- factor(ex$a, levels = ex$a[order(ex$b)])

ggplot(
    # uncomment this to see the other failure:
    # ex, aes(y = a, x = c)
) +
    geom_rect(
        aes(
            xmin = 'b',
            xmax = 'd',
            ymin = 'd',
            ymax = 'j'
            ),
        alpha = 0.2
    ) +
    geom_point(
        data = ex,
        aes(
            y = a,
            x = c
        )
    )

Answer 1

The best I can do is to avoid the data issue by using annotate instead (a data free layer). You somehow need to put geom_point first though, but I'm not sure why. It seems like the scale gets determined by the first layer, even though I have supplied the data and mapping already in ggplot.

ggplot(data = ex, aes(y = a, x = c)) +
    geom_point() +
    annotate(xmin = 'b', xmax = 'd', ymin = -Inf, ymax = Inf, geom = 'rect', alpha = 0.2)