Minor grid lines in ggplot2 with discrete values and facet grid - r

Question

I have a plot created using ggplot2 where I'm trying to modify some of the minor grid lines. Here is the current version:
library(tidyverse)
data(starwars)
starwars = starwars %>%
filter(!is.na(homeworld), !is.na(skin_color)) %>%
mutate(tatooine = factor(if_else(homeworld == "Tatooine", "Tatooine Native", "Other Native")),
skin_color = factor(skin_color))
ggplot(starwars, aes(birth_year, skin_color)) +
geom_point(aes(color = gender), size = 4, alpha = 0.7, show.legend = FALSE) +
facet_grid(tatooine ~ ., scales = "free_y", space = "free_y", switch = "y") +
theme_minimal() +
theme(
panel.grid.major.x = element_blank(),
panel.grid.major.y = element_blank(),
axis.title.x = element_blank(),
axis.title.y = element_blank(),
strip.placement = "outside",
strip.background = element_rect(fill="gray90", color = "white"),
) +
geom_hline(yintercept = seq(0, length(unique(starwars$skin_color))) + .5, color="gray30")
Y axis is a factor and a facet grid is used, with an uneven number of categories in each grid. I added some minor grid lines using geom_hline (my understanding is that panel.grid.minor does not work with categorical data i.e., factors).
I would like to remove the lines highlighted in yellow below, and then ADD a single black line in between the two facet grids (i.e., where the current double lines are that are highlighted in yellow).
Any way to do this? I'd prefer avoiding hard coding the position of any lines, in case the data change. Thanks.

Answer 1

Removing the top and bottom grid lines dynamically is relatively easy. You code the line positions in the data set based on the faceting groups and exclude the highest and lowest value, and plot the geom_hline with an xintercept inside the aes() statement. That approach is robust to changing the data (to see that this approach works if you change the data, comment out the # filter(!is.na(birth_year)) line below).
library(tidyverse)
library(grid)
data(starwars)
starwars = starwars %>%
filter(!is.na(homeworld), !is.na(skin_color)) %>%
mutate(tatooine = factor(if_else(homeworld == "Tatooine", "Tatooine Native", "Other Native")),
skin_color = factor(skin_color)) %>%
# filter(!is.na(birth_year)) %>%
group_by(tatooine) %>%
# here we assign the line_positions
mutate(line_positions = as.numeric(factor(skin_color, levels = unique(skin_color))),
line_positions = line_positions + .5,
line_positions = ifelse(line_positions == max(line_positions), NA, line_positions))
plot_out <- ggplot(starwars, aes(birth_year, skin_color)) +
geom_point(aes(color = gender), size = 4, alpha = 0.7, show.legend = FALSE) +
geom_hline(aes(yintercept = line_positions)) +
facet_grid(tatooine ~ ., scales = "free_y", space = "free_y", switch = "y") +
theme_minimal() +
theme(
panel.grid.major.x = element_blank(),
panel.grid.major.y = element_blank(),
panel.grid.minor.y = element_line(colour = "black"),
axis.title.x = element_blank(),
axis.title.y = element_blank(),
strip.placement = "outside",
strip.background = element_rect(fill="gray90", color = "white"),
)
print(plot_out)
gives
However, adding a solid between the facets without any hardcoding is difficult. There are some possible ways to add borders between facets (see here), but if we don't know whether the facets change it is not obvious to which value the border should be assigned. I guess there is a possible solution with drawing a hard coded line in the plot that divides the facets, but the tricky part is to determine dynamically where that border is going to be located, based on the data and how the facets are ultimately draw (e.g. in which order etc). I'd be interested in hearing other opinions on this.