Using position_stack() with text geom in stat_summary()

[aosmith16]

Should it be possible to "stack" text when calculating the labels and the y axis with stat_summary()? I'm getting an error using either position = "stack" or position = position_stack().

library(ggplot2)

# Using stat_summary for aggregation
ggplot(mtcars, aes(x = factor(cyl), y = mpg, fill = factor(am))) + 
    stat_summary_bin(fun.y = sum, geom="bar", position = "stack")+
    stat_summary(aes(label = stat(y)), fun.y = sum,  geom = "text", 
                 position = position_stack(vjust = .5))
#> Error in if (any(negative)) {: missing value where TRUE/FALSE needed

Things work as I expect when doing the calculation outside of ggplot2.

# Aggregation outside ggplot2
suppressPackageStartupMessages( library(dplyr) )
mtsum = mtcars %>%
    group_by(cyl, am) %>%
    summarise(mpg = sum(mpg) )

ggplot(mtcars, aes(x = factor(cyl), y = mpg, fill = factor(am))) + 
    stat_summary_bin(fun.y = sum, geom="bar", position = "stack")+
    geom_text(data = mtsum, aes(label = mpg), position = position_stack(vjust = .5))

Created on 2018-06-21 by the reprex package (v0.2.0).

[paleolimbot]

I think the problem is that PositionStack$compute_panel() assumes that there is a ymax column and that all of the values are non-missing:

ggplot2/R/position-stack.r

Line 179 in e2bdf85

negative <- data$ymax < 0

Not all geometries have a ymax and/or ymin (although most of the ones that are usually stacked do, like geom_bar(), which is why this isn't usually a problem). Changing this to something more like:

ymax_position <- ifelse(is.na(data$ymax), data$y, data$ymax)
negative <- ymax_position < 0

Would probably fix the problem. Given that collide() handles NA values in ymin and ymax, I think this is a good solution.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/