position_dodge2 with NA values causes errors

[espher1987]

I have one problem with geom_boxplot works. First I am using date type data and try not to plot all my data, only a part.

First I load my packages

library(dplyr)
library(ggplot2)
library(tibble)
library(quantmod)

I am using getSymbols to get my data

yahoo <- getSymbols(auto.assign = F,Symbols = "QQQ")
chartSeries(yahoo,subset = "last 3 month")

I shape my data to use it in ggplot

df <- yahoo %>%
  as.data.frame() %>% 
  rownames_to_column() %>% 
  rename("time" = "rowname") %>% 
  mutate(time = as.Date(time))

I set my date limits

l <- c(df$time[10],df$time[100])

Then I try to plot it

ggplot(data = df) +
  geom_boxplot(aes(x = time,
                   lower  = QQQ.Close,
                   upper  = QQQ.Open,
                   ymin   = QQQ.Low,
                   ymax   = QQQ.High,
                   middle = QQQ.Open),
               stat = "identity") +
  scale_x_date(limits = l)

and I get error:

Error in if (df$xmin[i] >= df$xmax[i - 1]) { :
valor ausente donde TRUE/FALSE es necesario

---

i get this error using ggplot2 3.0.0. I try this code in ggplot2 2.2.1 and i don't get any error. As far i search this can be related to "position_dodge2".

[batpigandme]

Could you please turn this into a reprex (short for minimal reproducible example)? It's especially helpful for ggplot2, since it automatically generates and uploads the plots (or lack thereof), which makes it much easier to go through at a glance.

If you've never heard of a reprex before, you might want to start by reading the tidyverse.org help page.

Thanks

[batpigandme]

If you're looking for a workaround, your code works if you filter on dates before plotting, and don't set limits in scale_x_date()

library(dplyr)
library(ggplot2)
library(tibble)
library(quantmod)
df <- yahoo %>%
  as.data.frame() %>% 
  rownames_to_column() %>% 
  rename("time" = "rowname") %>% 
  mutate(time = as.Date(time))

l <- c(df$time[10],df$time[100])

df_filtered <- df %>%
  filter(time >= l[1]) %>%
  filter(time <= l[2])

ggplot(data = df_filtered) +
  geom_boxplot(aes(x = time,
                   lower  = QQQ.Close,
                   upper  = QQQ.Open,
                   ymin   = QQQ.Low,
                   ymax   = QQQ.High,
                   middle = QQQ.Open),
               stat = "identity") +
  scale_x_date()

^{Created on 2018-09-25 by the reprex package (v0.2.1.9000)}

[paleolimbot]

I think this is a potential bug in position_dodge2() when xmin or xmax are NA.

library(ggplot2)
df <- data.frame(
  x_date = as.Date(c("1999-01-01", NA)),
  x_chr = c("char", NA),
  x_num = c(1, NA),
  y = 1
)
p <- ggplot(df, aes(y = y))

# these work
p + aes(xmin = x_date, xmax = x_date) + geom_errorbarh(position = "dodge")
#> Warning: Removed 1 rows containing missing values (geom_errorbarh).

p + aes(xmin = x_chr, xmax = x_chr) + geom_errorbarh(position = "dodge2")

# these fail
p + aes(xmin = x_date, xmax = x_date) + geom_errorbarh(position = "dodge2")
#> Error in if (df$xmin[i] >= df$xmax[i - 1]) {: missing value where TRUE/FALSE needed
p + aes(xmin = x_num, xmax = x_num) + geom_errorbarh(position = "dodge2")
#> Error in if (df$xmin[i] >= df$xmax[i - 1]) {: missing value where TRUE/FALSE needed

^{Created on 2019-05-24 by the reprex package (v0.2.1)}

In the original dataset for this issue there weren't any NAs until the values were censored by scale_x_date() with limits. The error is being generated here:

ggplot2/R/position-dodge2.r

Lines 132 to 143 in 1f6f0cb

	find_x_overlaps <- function(df) {
	overlaps <- numeric(nrow(df))
	overlaps[1] <- counter <- 1
	for (i in seq_asc(2, nrow(df))) {
	if (df$xmin[i] >= df$xmax[i - 1]) {
	counter <- counter + 1
	}
	overlaps[i] <- counter
	}
	overlaps
	}