Modify global data from within a function in R - r

Question

I am trying to write a function that will add a new column to a data frame, when I call it, without doing any explicit assignment.
i.e I just want to call the function with arguments and have it modify the data frame:
input_data:
x y
1 2
2 6
column_creator<-function(data,column_name,...){
data$column_name <- newdata ...}
column_creator(input_data,new_col,...)
x y new_col
1 2 5
2 6 9
As opposed to:
input_data$new_col <- column_creator(input_data,new_col,...)
However doing assignment inside the function is not modifying the global variable.
I am working around this by having the function return a statement of assignment (temp in the function below), however is there another way to do this?
Here is my function for reference, it should create a column of 1s inbetween the supplied start and end date with the name dummy_name.
dummy_creator<-function(data,date,dummy_name,start,end){
temp<-paste(data,"['",dummy_name,"'] <- ifelse(",data,"['",date,"'] > as.Date (","'" , start,"'" , ", format= '%Y-%m-%d') & ",data,"['",date,"'] < as.Date(", "'", end,"'" ,",format='%Y-%m-%d') ,1,0)",sep="")
print(temp)
return()
}
Thanks
I also tried:
dummy_creator<-function(data,date,dummy_name,start,end){
data[dummy_name] <<- ifelse(data[,date] > as.Date (start, format= "%Y-%m-%d") & data[,date] < as.Date(end,format="%Y-%m-%d") ,1,0)
}
But that attempt gave me error object of type closure is not subsettable.

Answer 1

It’s generally a bad idea to modify global data or data passed into a function: R objects are immutable, and using tricks to modify them inside a function breaks the user’s expectations and makes it harder to reason about the program’s state.
It is good form to return the modified object instead:
input_data = column_creator(input_data, new_col, …)
That said, you have a few options. Generally, R has several mechanisms to allow modifiable objects. I recommend you look into R6 classes for this.
You could also use non-standard evaluation to capture the passed object and modify it at the caller’s site. However, this is rarely advisable. I’m posting an example of this here because the mechanism is interesting and worth knowing, but I’ll reiterate that you shouldn’t use it here.
function (df, new_col, new_data) {
# Get the unevaluated expression representing the data frame
df_obj = substitute(df)
new_col = substitute(new_col)
# Ensure that the input is valid
stopifnot(is.name(df_obj))
stopifnot(is.name(new_col))
stopifnot(is.data.frame(df))
# Add new column to data frame
df[[deparse(new_col)]] = new_data
# Assign back to object in caller scope
assign(deparse(df_obj), df, parent.frame())
invisible(df)
}
test = data.frame(A = 1 : 5, B = 1 : 5)
column_creator(test, C, 6 : 10)
test
# A B C
# 1 1 1 6
# 2 2 2 7
# 3 3 3 8
# 4 4 4 9
# 5 5 5 10