Can't save an object within function

Question

I'm stuck and can't for the heck of it figure out what I'm missing. I'm sure it's something minor, but I just really don't understand why my code isn't working.

myfunction <- function( X ) {

plot(X, type="l")
max_B <- max(X$B)   
max_A <- X$A[ X$B  == max_B ]
results<-c(max_A, max_B)
points(max_A, max_B, col="red")  

} 

myfunction(dat)

It's plotting just fine, it's adding the point at max_A/max_B (so it's actually getting information for max_A and max_B, but it's not creating the object results. If I do every line within the function separately, it works, but it just doesn't work as a function and I have no idea why.

Answer 1

you might need to return the object.

myfunction <- function( X ) {
  plot(X, type="l")
  max_B <- max(X$B)   
  max_A <- X$A[ X$B  == max_B ]
  results<-c(max_A, max_B)
  points(max_A, max_B, col="red")  
  return(results)
} 

results <- myfunction(dat)

In case this doesn't solve the problem, please provide date for dat by providing the result of dput(dat) in your question.

You don't need always the return in the function. The default setting of functions is to return the last saved object or last instructuion. In your case, the points function has no or invisible return or at least not the one you expect. I think for clean code practice it is always good to define the return value, even if it is one more line to write.

Answer 2

I wanted to provide some feedback in case someone else comes across this. The right answer was actually posted by drmariod, but I missed the most important part. I not only have to return my object, but actually save the returned output into a new object when executing the function. Which it says right there - I just totally overlooked that part.

myfunction<-function()
{...
return(object)}

results<-myfunction(dat)

Thank you again for your help.

Answer 3

Push the assignment of results one more level above to your global environment should solve your problem.

myfunction <- function( X ) {

  plot(X, type="l")
  max_B <- max(X$B)   
  max_A <- X$A[ X$B  == max_B ]
  results <<- c(max_A, max_B)
  points(max_A, max_B, col="red")  

} 

myfunction(dat)

For more information of <<-, check How do you use “<<-” (scoping assignment) in R?, and also the Environment chapter in Advanced R written by Hadley Wickham. You'll learn more about how R scopes variables across environments.

Why isn't the normal return working?

We may return local variables out to Global Environment, but remember that returns only the values of returned local variables, but not the local variables themselves.

For example, if we create a function simply adds one to a given number:

add1 <- function(x) {
  y <- x + 1
  return(y)
  # return(x + 1) will also do but I save it as y for better demonstration
}

add1(1) gives us 2 on the screen, but doesn't give us a variable called "y" in Global Environment. If we really want an y out there, we have to save the results from add(1) and name it as "y", i.e. y <- add1(1). Instead of showing "2" on the screen, now it is saved as y.

The trick of <<- is pushing the assignment to its parent environment. In this answer, results <<- c(max_A, max_B) takes place in myfunction, whose parent environment happens to be Global Environment, so we'll find results after executing myfunction.

Why can I use the objects for to plot the point, but not to save the very same objects into a new object?

The objects max_A and max_B are effective locally in myfunction, so you can do whatever you want within that scope. In your original codes, results was indeed saved but still existed only in myfunction environment.

One thing worth mentioning is that, unlike most other functions, plot and points returns no object but graphics, so they are not limited by the environment scopes. You can try x <- 1:10; y <- x + rnorm(10); myplot <- plot(x, y), and you'll see a plot but myplot is empty.