Remove baseline color for geom_histogram - r

Question

I'm adding a color aesthetic to a faceted histogram. In the reprex below, with no color aesthetic, the histogram only show data within that facet level. However, with color defined, a baseline is added which stretches the stretches to include the range of data across all facets. Is there a way to make this not happen?
I'm looking for something similar to geom_density with trim = TRUE, but there doesn't appear to be a trim option for geom_histogram.
library(tidyverse)
data <- tibble(a = rchisq(1000, df = 3),
b = rchisq(1000, df = 1),
c = rchisq(1000, df = 10)) %>%
gather()
ggplot(data, aes(x = value)) +
geom_histogram() +
facet_wrap(~ key, ncol = 1)
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
ggplot(data, aes(x = value)) +
geom_histogram(color = "red") +
facet_wrap(~ key, ncol = 1)
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
ggplot(data, aes(x = value)) +
geom_density(color = "red", trim = TRUE) +
facet_wrap(~ key, ncol = 1)
Created on 2019-07-20 by the reprex package (v0.3.0)

Answer 1

geom_histogram draws its bars using using rectGrob from the grid package, and a zero-width / zero-height rectGrob is depicted as a vertical / horizontal line in the outline colour, at least in my set-up for RStudio (& OP's as well, I presume). Demonstration below:
library(grid)
r1 <- rectGrob(width = unit(0, "npc"), gp = gpar(col = "red", fill = "grey")) # zero-width
r2 <- rectGrob(height = unit(0, "npc"), gp = gpar(col = "red", fill = "grey")) # zero-height
grid.draw(r1) # depicted as a vertical line, rather than disappear completely
grid.draw(r2) # depicted as a horizontal line, rather than disappear completely
In this case, if we check the data frame associated with the histogram layer, there are many rows with ymin = ymax = 0, which are responsible for the 'baseline' effect seen in the question.
p <- ggplot(data, aes(x = value)) +
geom_histogram(color = "red") +
facet_wrap(~ key, ncol = 1)
View(layer_data(p) %>% filter(PANEL == 2)) # look at the data associated with facet panel 2
Workaround: Since the data calculations are done in StatBin's compute_group function, we can define an alternative version of the same function, with an additional step to drop the 0-count rows from the data frame completely:
# modified version of StatBin2 inherits from StatBin, except for an
# additional 2nd last line in compute_group() function
StatBin2 <- ggproto(
"StatBin2",
StatBin,
compute_group = function (data, scales, binwidth = NULL, bins = NULL,
center = NULL, boundary = NULL,
closed = c("right", "left"), pad = FALSE,
breaks = NULL, origin = NULL, right = NULL,
drop = NULL, width = NULL) {
if (!is.null(breaks)) {
if (!scales$x$is_discrete()) {
breaks <- scales$x$transform(breaks)
}
bins <- ggplot2:::bin_breaks(breaks, closed)
}
else if (!is.null(binwidth)) {
if (is.function(binwidth)) {
binwidth <- binwidth(data$x)
}
bins <- ggplot2:::bin_breaks_width(scales$x$dimension(), binwidth,
center = center, boundary = boundary,
closed = closed)
}
else {
bins <- ggplot2:::bin_breaks_bins(scales$x$dimension(), bins,
center = center, boundary = boundary,
closed = closed)
}
res <- ggplot2:::bin_vector(data$x, bins, weight = data$weight, pad = pad)
# drop 0-count bins completely before returning the dataframe
res <- res[res$count > 0, ]
res
})
Usage:
ggplot(data, aes(x = value)) +
geom_histogram(color = "red", stat = StatBin2) + # specify stat = StatBin2
facet_wrap(~ key, ncol = 1)

Answer 2

Another option could be using after_stat on your y aes with an ifelse to check if the mapped value is higher than 0 otherwise replace the value with NA which will make it possible to remove the baseline color like this:
library(tidyverse)
ggplot(data, aes(x = value, y = ifelse(after_stat(count) > 0, after_stat(count), NA))) +
geom_histogram(color = "red") +
facet_wrap(~ key, ncol = 1)
Created on 2023-02-15 with reprex v2.0.2