boxplot computation fails when setting scale limits

[cpsievert]

Before 10fa001, this code produced the following image:

ggplot(PlantGrowth, aes(y = weight)) +
  geom_boxplot() +
  scale_y_continuous(limits = c(5, 7.5))

After 10fa001, it now produces a blank plot with this warning:

Warning message:
Computation failed in `stat_boxplot()`:
missing values and NaN's not allowed if 'na.rm' is FALSE

cc @thomasp85

[rudeboybert]

I believe the issue is more fundamental than scale limits. Consider the following code:

library(ggplot2)
library(nycflights13)

ggplot(data = weather, mapping = aes(x = factor(month), y = temp)) +
  geom_boxplot()

Using ggplot2_3.2.1 yields

whereas using ggplot2_3.2.1.9000 yields

along with the same error message as above:

#> Warning: Computation failed in `stat_boxplot()`:
#> missing values and NaN's not allowed if 'na.rm' is FALSE

cc @ismayc

[clauswilke]

I think I found the problem. It's in this line:

ggplot2/R/stat-boxplot.r

Line 55 in 10fa001

vars = "x",

It should read vars = "y".

Reprex:

library(ggplot2)
library(rlang)
library(nycflights13)

# doesn't work
ggplot(data = weather, mapping = aes(x = factor(month), y = temp)) +
  geom_boxplot()
#> Warning: Computation failed in `stat_boxplot()`:
#> missing values and NaN's not allowed if 'na.rm' is FALSE

# fix setup_data() function of StatBoxplot
StatBoxplot$setup_data <- function(data, params) {
  data <- ggplot2:::flip_data(data, params$flipped_aes)
  data$x <- data$x %||% 0
  data <- ggplot2:::remove_missing(
    data,
    na.rm = FALSE,
    vars = "y",
    name = "stat_boxplot"
  )
  flip_data(data, params$flipped_aes)
}

# now it works
ggplot(data = weather, mapping = aes(x = factor(month), y = temp)) +
  geom_boxplot()
#> Warning: Removed 1 rows containing missing values (stat_boxplot).

^{Created on 2019-10-06 by the reprex package (v0.3.0)}

cc @thomasp85

[clauswilke]

Thomas, while we're at it, is this line correct, or should it read na.rm = params$na.rm?

ggplot2/R/stat-boxplot.r

Line 54 in 10fa001

na.rm = FALSE,

It looks like with the current settings it's impossible to suppress the warning about missing values.

library(ggplot2)
library(rlang)
library(nycflights13)

# fix setup_data() function of StatBoxplot
StatBoxplot$setup_data <- function(data, params) {
  data <- ggplot2:::flip_data(data, params$flipped_aes)
  data$x <- data$x %||% 0
  data <- ggplot2:::remove_missing(
    data,
    na.rm = FALSE,
    vars = "y",
    name = "stat_boxplot"
  )
  flip_data(data, params$flipped_aes)
}

# warning doesn't get suppressed
ggplot(data = weather, mapping = aes(x = factor(month), y = temp)) +
  geom_boxplot(na.rm = TRUE)
#> Warning: Removed 1 rows containing missing values (stat_boxplot).

StatBoxplot$setup_data <- function(data, params) {
  data <- ggplot2:::flip_data(data, params$flipped_aes)
  data$x <- data$x %||% 0
  data <- ggplot2:::remove_missing(
    data,
    na.rm = params$na.rm,
    vars = "y",
    name = "stat_boxplot"
  )
  flip_data(data, params$flipped_aes)
}

# now the warning gets suppressed
ggplot(data = weather, mapping = aes(x = factor(month), y = temp)) +
  geom_boxplot(na.rm = TRUE)

^{Created on 2019-10-06 by the reprex package (v0.3.0)}

[thomasp85]

@clauswilke I don't know if it was intended that way (hardcoding the na.rm). @karawoo do you remember.

As for the cause of the original issue, it is actually much more complicated and is due to the new x|y notation in required_aes - will figure out a fix

[karawoo]

I think hardcoding na.rm was just an oversight, I'll fix it

[rudeboybert]

Thank you for fixing this so promptly. I would like to give credit to my student Nouhaila @nouny for finding the error with the temp vs month boxplot.

[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/