Format geom_text label doesn't work when using aes_string

Question

I am using the dot function to format text labels in a plot created with ggplot2. This works fine when using aes, but doesn't work like expected when using aes_string. Is there a workaround to make it work with aes_string?

require(ggplot2)

# Define the format function
dot <- function(x, ...) { 
  format(x, ..., big.mark = ".", scientific = FALSE, trim = TRUE)
}

# Create dummy data
df <- data.frame(cbind(levels(iris$Species),c(10000000000,200000,30000)))
df$X2 <- as.numeric(as.character(df$X2))

# Works with aes   
ggplot(iris) + 
  geom_bar(aes(Species,Sepal.Width),stat="identity") +     
  geom_text(data=df,aes(x=factor(X1),y=180,label=dot(X2)))


# Doesn't work with aes_string
ggplot(iris) + 
  geom_bar(aes(Species,Sepal.Width),stat="identity") +
  geom_text(data=df,aes_string(x="X1",y=180,label=dot("X2")))

I think the easiest thing would be to simply apply `dot` outside of ggplot, creating a new column. — joran, Jun 18 '14 at 18:50

Anders Ellern Bilgrau · Answer 1 · 2014-06-18T19:06:34.867

3

You're passing a string constant to dot and dot("X2") returns "X2". So you're basically giving aes_string the argument label = "X2".

I got this to work

ggplot(iris) + 
  geom_bar(aes(Species, Sepal.Width), stat = "identity") + 
  geom_text(data=df, aes_string(x="X1", y =180, label = deparse(dot(df$X2))))

Imgur

Edit

As suggested by MrFlick, you could also use df[,"X2"] if you need to pass the column name as a string. I.e.

ggplot(iris) + 
  geom_bar(aes(Species, Sepal.Width), stat = "identity") + 
  geom_text(data=df, aes_string(x="X1", y =180, label = deparse(dot(df[,"X2"]))))

edited Jun 18 '14 at 19:06

answered Jun 18 '14 at 18:54

Anders Ellern Bilgrau

9,928
1
30
37

In order to specify "X2" as a string, you could use `aes_string(x="X1", y =180, label = deparse(dot(df[,"X2"]))))` – MrFlick Jun 18 '14 at 18:59

MrFlick · Accepted Answer · 2014-06-18T19:15:30.057

3

Rather than just quote "X2", you must quote the whole expression

ggplot(iris) + 
  geom_bar(aes(Species, Sepal.Width), stat = "identity") + 
  geom_text(data=df, aes_string(x="X1", y =180, label = "dot(X2)"))

If you wanted to specify variable names via a character vector, you would use paste() to build that expression.

edited Jun 18 '14 at 19:15

answered Jun 18 '14 at 19:07

MrFlick

195,160
17
277
295

Arh, that's the way to do it! Perhaps you should use `aes_string` and not `aes_foo` to have actual runable code? – Anders Ellern Bilgrau Jun 18 '14 at 19:11
Thanks @AEBilgrau I was testing some code to see what `aes_string` actually did and forgot to take that test function out. – MrFlick Jun 18 '14 at 19:16

score 2 · Answer 3 · answered Jun 18 '14 at 19:24

As implied by my comment, I dislike relying on deparse or other evaluation tricks unless I have to:

col <- "X2"
new_col <- paste0(col,"_dot")
df[,new_col] <- dot(df[,col])
ggplot(iris) + 
    geom_bar(aes(Species,Sepal.Width),stat="identity") + 
    geom_text(data=df,aes_string(x="X1",y=180,label=new_col))

I presented this version as if we were writing code in a function and had received the "X2" as an argument.

Format geom_text label doesn't work when using aes_string

3 Answers3

Edit

Linked