Writing plotting functions with the ability to choose by name which columns will be plotted

[Roey Angel]

Hi,
I'm trying to generate some ggplot functions but give the user the option of choosing which variables to plot by parsing their names (data frame column names).
Seems like a basic option to me to enable function writing but so far I've been unsuccessful -- ggplot either gives an error or otherwise funny results.
For example, given the following minimal data frame:

library(ggplot2)
data2plot <- data.frame(x1 = seq(1,10), y1 = seq(1,10), y2 = seq(10,1))

Now I'd like to generate a ggplot function that will plot a simple scatter plot but will accept column names as input so that the user could choose by name which column to plot.
but...

Running this:

plot.variable <- function(data2plot, y.axis){
 qplot(x1, y = y.axis, data = data2plot)
}
plot.variable(data2plot, y.axis = 'y1')

Plots ten points on the graph with the Y value = 'y1'

Running this:

plot.variable <- function(data2plot, x.axis, y.axis){
  qplot(x = x.axis, y = y.axis, data = data2plot)
}
plot.variable(data2plot, x.axis = 'x1', y.axis = 'y1')

Plots one point 'x1' vs 'y1'
     
Running this:

plot.variable <- function(data2plot, x.axis, y.axis){
  qplot(data = data2plot, x = data2plot[, x.axis], y = data2plot[, y.axis])
}
plot.variable(data2plot, x.axis = 'x1', y.axis = 'y1')

Actually gives a plot but if I write a function using 'ggplot' instead of qplot:

plot.variable <- function(data2plot, x.axis, y.axis){
  ggplot(data2plot, aes(x = data2plot[, x.axis], y = data2plot[, y.axis])) +
    geom_point()
}
plot.variable(data2plot, x.axis = 'x1', y.axis = 'y1')

Gives and error: 'Error in `[.data.frame`(data2plot, , x.axis) : object 'x.axis' not found'
I could get this function to not give an error if I assign the variables outside the function call (i.e. x.axis <- 'x1', y.axis <- 'y1'), but that causes a mess with more complex functions which include faceting etc. I guess because it doesnt keep the original order of observations as in the data frame. Besides that's obviosly not a solution to my problem.

Is there a solution to this?

Thanks in advance,
Roey

[Winston Chang]

If you use aes_string() instead of aes(), you can do it.

Notice how these give the same result:

> aes(x = weight)

List of 1

 $ x: symbol weight

> xname <- "weight"

> aes_string(x = xname)

List of 1

 $ x: symbol weight

-Winston

--
You received this message because you are subscribed to the ggplot2 mailing list.
Please provide a reproducible example: https://github.com/hadley/devtools/wiki/Reproducibility
 
To post: email ggp...@googlegroups.com
To unsubscribe: email ggplot2+u...@googlegroups.com
More options: http://groups.google.com/group/ggplot2

[Roey Angel]

Thanks Winston,
I actually knew about aes_string but recall having trouble with it in the past as well.
In any case since I cannot reproduce the problem I think I had in the past and since the code works now I'll stick to it.

Thanks again
Roey