djhurio
djhurio

Reputation: 5536

Object not found error with ggplot2

I can not get my head around this.

These examples are working:

# Function with geom_density

gr.den <- function(var.name) {
  ggplot(results, aes(get(var.name), fill = name)) +
  geom_density(alpha = 0.2) +
  geom_vline(xintercept = tv[, var.name], color="red", size=1) +
  xlab(var.name)
}

gr.den("sum.Empl")

# Example with geom_point

ggplot(results, aes(sum.All, sum.Empl)) +
  geom_point(alpha = 1/5) +
  opts(aspect.ratio = 1) +
  facet_grid(. ~ name)

Then I am trying to create similar function using geom_point:

gr.sc <- function(var.name.1, var.name.2) {
  ggplot(results, aes(get(var.name.1), get(var.name.2))) +
  geom_point(alpha = 1/5) +
  opts(aspect.ratio = 1) +
  facet_grid(. ~ name)
}

gr.sc("sum.All", "sum.Empl")

And I am getting this error. Why?

Error in get(var.name.1) : object 'var.name.1' not found

Upvotes: 9

Views: 26279

Answers (3)

user3409049
user3409049

Reputation: 11

I ran into problems similar to this: https://groups.google.com/forum/#!topic/ggplot2/_kKP4NNu3bc where aes_string does not work.

This might be resolved in later versions of ggplot2. But I'm using an older version of ggplot2 for compatibility reasons. A quick hack that worked for me was to set function parameters as global variables, i.e.

gr.sc <- function(var.name.1, var.name.2) {
    var.name.1 <<- var.name.1
    ...
    ggplot(...
    ...
}

Upvotes: 1

Luciano Selzer
Luciano Selzer

Reputation: 10016

If you are going to use aes inside a function it's better to use aes_string instead.

gr.sc <- function(var.name.1, var.name.2) {
  ggplot(results, aes_string(x = var.name.1, y = var.name.2)) +
  geom_point(alpha = 1/5) +
  opts(aspect.ratio = 1) +
  facet_grid(. ~ name)
}

gr.sc("sum.All", "sum.Empl")

HTH

Upvotes: 18

Richie Cotton
Richie Cotton

Reputation: 121137

The error occurs because get is looking in the wrong environment (i.e., not inside the results data frame). You could explicitly specify the get(var.name.1, envir = results) but that would be ugly, awful code. Much better to use aes_string as Iselzer suggests.

Upvotes: 5

Related Questions