Here is a situation where using cur_data() and across() yields the same result:
library(tidyverse)
by_cyl <- mtcars %>% group_by(cyl)
a <- summarise(by_cyl,head(cur_data(), 2))
b <- summarise(by_cyl,head(across(), 2)))
identical(a,b)
#TRUE
My question is what is the difference between these two functions? When are they interchangeable and when are they not?
I think there is some misunderstanding. across and cur_data() are not related/interchangeable because they both are for different purpose.
across is used to apply a function on multiple columns, cur_data returns the current data for a group without the grouping variable.
In the example shared when you are using across() it does nothing and returns data as it is.
For example,
library(dplyr)
mtcars %>% summarise(across())
returns the same mtcars dataset as it is because the default values in across are .cols = everything() and .fns = NULL. So it applies NULL function to all the columns.
I am not sure what output you were expecting in this case.