Object created inside function not found by ggplot

Question

I have a csv of time series data for a number of sites that I produce ggplots for, showing changes in means using the changepoint package. I have written a function that takes the csv, performs some calculations to get the means then loops through the sites producing a plot for each. My problem is that an object created in the for loop isn't found.

A very simplified example is below but produces the same error:

df1 <- data.frame(date = seq(as.Date("2015-01-01"), as.Date("2015-01-10"),
                         by = "day"),
              site1 = runif(10),
              site2 = runif(10),
              site3 = runif(10))

example <- function(df1){

    sname <- names(df1)[-1]

    for (i in 1:length(sname)){
            df2 <- df1[,c(1, 1+i)]
            df2$label <- factor(rep("ts", by=length(df2[,1])))

            plot <- ggplot()+
                    geom_point(data=df2, aes(x=date, y=df2[,2]))+
                    geom_line(data=df2, aes(x=date, y=df2[,2]))
            sname.i<-sname[i]
            filename<-paste0(sname.i, "-test-plot.pdf")
            ggsave(file=filename, plot)
    }
}

example(df1)

The error I get is: " Error in eval(expr, envir, enclos) : object 'df2' not found"

I'm not quite sure what the problem is as I have created similar loops which have worked in the past. If I assign a value to i and step through the code within the loop it works fine. I'm thinking an environment problem? Or is ggsave doing something wiggy? Any help/pointers gratefully received. Thanks.

When I run your code, the problem is sname. Without this variable being defined, i also remains undefined, and the same then happens to df2. — vaettchen
– vaettchen, Commented May 22, 2015 at 8:24
@vaettchen. Did you run the whole function or just the loop? I re-ran the example function on my home computer and still get the error as stated above. It fails for me on df2. — Bart
– Bart, Commented May 22, 2015 at 9:45

Joris Meys · Accepted Answer · 2015-05-22 12:34:58Z

4

You problem is not so much your code, but the implementation of the ggplot2 package. This package uses nonstandard evaluation, and that can seriously mess up your results.

Take a look at the example code at the end of this post. I create in the global environment a data frame called df2 with different values. If I run your code now, you get plots that looks like this:

enter image description here

Note that on the X axis, it uses the correct dates, but the values on the Y axis are the ones from the dataframe df2 that is in the global environment! So the function aes() looks for the data in two different places. If you specify the name of a variable as a symbol (date) the function first looks in the data frame that is specified in the function call. However, an expression like df2[,2] cannot be found in the dataframe, as there is no variable with that name. Due to the way the ggplot2 package is constructed, R will look for that in the global environment instead of the calling environment.

As per wici's comment: Your best option is probably to use the function aes_string(), as this allows you to pass the aes in character form, and this function evaluates expressions in the correct environment :

plot <- ggplot()+
      geom_point(data=df2, aes_string(x="date", y=sname[i]))+
      geom_line(data=df2, aes_string(x="date", y=sname[i]))

Alternatively, you can get around that by using eval() and parse() like this:

example <- function(df1){

  sname <- names(df1)[-1]

  for (i in 1:length(sname)){
    df2 <- df1[,c(1, 1+i)]
    df2$label <- factor(rep("ts", by=length(df2[,1])))

    aesy <- sname[i]
    command <- paste("plot <- ggplot()+
      geom_point(data=df2, aes(x=date, y=",aesy,"))+
      geom_line(data=df2, aes(x=date, y=",aesy,"))")

    eval(parse(text=command))                     
    sname.i<-sname[i]
    print(plot)
  }

If you try that out with the example script below, you'll see that this time around you get the correct values displayed. Note that this is a suboptimal solution, as most solutions involving eval(). I'd go for aes_string() here.

EXAMPLE SCRIPT

df1 <- data.frame(date = seq(as.Date("2015-01-01"), as.Date("2015-01-10"),
                             by = "day"),
                  site1 = runif(10),
                  site2 = runif(10),
                  site3 = runif(10))

df2 <- data.frame(date = seq(as.Date("2014-10-01"), as.Date("2014-10-10"),
                             by = "day"),
                  site1 = runif(10,10,20),
                  site2 = runif(10,10,20),
                  site3 = runif(10,10,20))

example <- function(df1){

  sname <- names(df1)[-1]

  for (i in 1:length(sname)){
    df2 <- df1[,c(1, 1+i)]
    df2$label <- factor(rep("ts", by=length(df2[,1])))

    plot <- ggplot()+
      geom_point(data=df2, aes(x=date, y=df2[,2]))+
      geom_line(data=df2, aes(x=date, y=df2[,2]))

    sname.i<-sname[i]
    print(plot)
  }
}

example(df1)

edited May 22, 2015 at 12:34

answered May 22, 2015 at 10:17

Joris Meys

109k31 gold badges228 silver badges266 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

wici Over a year ago

You are of course right, but ggplot2 also provides a function for regular evaluation of aesthetics. Just use aes_string(x="date", y=aesy) as your aesthetic generator.

Bart Over a year ago

@Joris. Thanks for that. In my efforts to convert what I used in base graphics to ggplot I can see where I've been going wrong.

Joris Meys Over a year ago

@wici Thanks for pointing that out! I wasn't aware of that function aes_string. I've added it to the answer. If you want to post it as your own answer, let me know and I remove it again. You deserve to get the full credit for this.

Collectives™ on Stack Overflow

Object created inside function not found by ggplot

1 Answer 1

3 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

3 Comments

Your Answer

Sign up or log in

Post as a guest

Related