Refactoring `<<-`: Thread on function-specific and package-specific environments

If you are curious about the use of private environments for functions and packages, have a read of this nice R-package-devel thread!

https://www.mail-archive.com/r-package-devel@r-project.org/msg07413.html

This offers some ideas we may use to easily refactor <<- in our code so that we touch the least code and make it substantially safer. The entire thread is worth reading but one notable way is to wrap our code in local().

# Define a variable in the global environment
variable <- 0
set_variable <- function(x) {
  variable <<- 99
  variable
}

# set_variable() dangerously overwrites the `variable` in the global environment
variable <- 1
set_variable()
#> [1] 99
variable == 1
#> [1] FALSE

set_variable2 <- local({
  variable <- 0
  function(x) {
    variable <<- 99
    variable
  }
})

# set_variable2() no longer overwrites the `variable` in the global environment
variable <- 1
set_variable2()
#> [1] 99
variable == 1
#> [1] TRUE

# Of course you can use set_variable2() inside other funcitons
use_variable <- function(x) {
  variable <- set_variable2()
  x + variable
}

use_variable(1)
#> [1] 100

^{Created on 2021-11-30 by the reprex package (v2.0.1)}

Read the linked R-package-devel thread first, focusing on its discussion of function-specific and package-specific environments and the use of local(). The issue names no files or tests, so the next step is to identify the current uses of <<- in the project and define which ones should be refactored. Done means a safer, agreed-upon refactoring with no unintended global-environment changes.