一个简单的可重现示例，在 ZE1E1D3D405731837E 中的自定义 function 中将 arguments 传递给 data.table

Question

I have been googling this answer for a few hours. A lot of people have asked similar questions, but I did not find either a simple enough question or a straightforward answer. Here is my approach:

Assume that I want to do a simple group by in data.table :

library(data.table)
mtcars = data.table(mtcars)
mtcars[,sum(mpg), gear]

# Here are the results
#   gear    V1
#1:    4 294.4
#2:    3 241.6
#3:    5 106.9

However, if I use a self-defined function to do this:

zz = function(data, var, group){
  return(data[,sum(var), group])
}
zz(mtcars, mpg, gear)

I got an error message:

Error in eval(bysub, parent.frame(), parent.frame()): object 'gear' not found

I've tried substitute , eval , quote , and other solutions, but none of them works. I wonder if anyone could give a more straightforward solution and explanation to this.

Thank you and happy Halloween!

Answer 1

If we are using unquoted arguments, substitute and eval uate

zz <- function(data, var, group){
 var <- substitute(var)
 group <- substitute(group)
 setnames(data[, sum(eval(var)), by = group],
        c(deparse(group), deparse(var)))[]
 # or use
 #  setnames(data[, sum(eval(var)), by = c(deparse(group))], 2, deparse(var))[]

}
zz(mtcars, mpg, gear)
#   gear   mpg
#1:    4 294.4
#2:    3 241.6
#3:    5 106.9

Answer 2

While not perfect, the ... argument can be helpful:

zz = function(dt, ...){
  return(dt[...])
}

zz(mtcars, , sum(mpg), gear)

   gear    V1
1:    4 294.4
2:    3 241.6
3:    5 106.9

Answer 3

I don't really see the point in writing a function that takes unquoted arguments. Why not just use the data.table syntax directly?

If you want to write a function it makes more sense to take a character vector of column names since this is way more programmable than symbols (think about programming with setkey versus setkeyv or writing functions to create ggplots with aes versus aes_string ) The downside is that the internal of the function is messy and requires eval(parse(text=.))) NSE in order for GForce to work correctly, but the function interface is more extensible.

zz = function(data, var, group){
  eval(parse(text=paste0("data[,sum(",var,"),by=",group,"]")))
}
zz(mtcars, "mpg", "gear")