[R] Using data and subset arguments in user-defined functions

[Duncan Murdoch]

On 10/27/2006 5:18 PM, Manuel Morales wrote:
> Dear list,
> 
> A while ago, I posted a question asking how to use data or subset
> arguments in a user-defined function. Duncan Murdoch suggested the
> following solution in the context of a data argument:
> 
> data <- data.frame(a=c(1:10),b=c(1:10))
> 
> eg.fn <- function(expr, data) {
>                   x <- eval(substitute(expr), envir=data)
>                   return(mean(x))
> }
> 
> eg.fn(a,data)
> 
> I've tried various approaches to add a subset argument to the example
> above, but no luck. I'm looking for something like the following (but
> that works!)
> 
> eg.fn2 <- function(expr, data, subset) {
>                    data <- subset(data,subset)
>                    x <- eval(substitute(expr), envir=data)
>                    return(mean(x))
> }
> 
> eg.fn2(a, data, subset=a>3)
> 
> This returns the error: 
> "Error in eg.fn2(a, data, subset = a > 3) : 
>         object "a" not found"
> 
> Any suggestions?

The problem is that subset needs to be evaluated with envir=data too. 
So this works:

eg.fn2 <- function(expr, data, subset) {
	subset <- eval(substitute(subset), envir=data)
  	data <- subset(data,subset)
         x <- eval(substitute(expr), envir=data)
         return(mean(x))
}

It doesn't work to skip the first assignment and just put it inline into

subset(data, eval(substitute(subset), envir=data))

because subset() does nonstandard evaluation of the second argument.

Duncan Murdoch