Passing in unbound variables into R functions

This is a case of the R-specific non-standard evaluation. This is a quite powerful concept in R and basically means that what you pass in the function select is not evaluated directly. Rather, it takes the argument unevaluated and evaluates it later, in the context of the dataframe.

I suggest you read the Advanced R chapter on this http://adv-r.had.co.nz/Computing-on-the-language.html.

To further expand on this, look at this example which showcases the basic concept:

expample_dataframe <- data.frame(
  foo = c(1:5),
  bar = c(10:14)
)

foo <- c("any" , "variable", "in", "global", "namespace")

print(foo) 
#> [1] "any"       "variable"  "in"        "global"    "namespace"

select_column <- function(data_frame, column_name){
  column_name <- substitute(column_name)
  eval(column_name, envir = data_frame)
}

select_column(expample_dataframe, foo)
#> [1] 1 2 3 4 5

foo
#> [1] "any"       "variable"  "in"        "global"    "namespace"

Explanation

substitute() leaves the input of the function unevaluated, i.e. it references it as a symbol. Then we can use the eval() function which allows you to evaluate a certain call/symbol inside a certain namespace. And yes, in R dataframes are namespaces. Thus

eval(column_name, envir = data_frame) evaluates column_name in the context of the dataframe.

This is what goes on behind the scenes of many functions in R.

^{Created on 2020-08-14 by the reprex package (v0.3.0)}