fixGlobals

A helper tool for package developers to avoid R CMD check notes concerning “no visible binding for global variable” that occur when a function refers to a variable that isn’t defined in the global environment.

These notes exist for obvious good reason but can become a nuisance when using data transformation packages like data.table and dplyr which make use of Non Standard Evaluation (NSE) to refer to columns in a data frame like R object. Every use of NSE creates an R CMD check note due to the fact that the R CMD check interprets unquoted variable names as global variables.

fixGlobals works by maintaining a character vector of variable names, passed to utils::globalVariables(), and which are ignored during the R CMD check.

How to use

Initiate zzz_global_variables.R

First the file “R/zzz_global_variables.R” must be initiated with fixGlobals::initiate_file(). This is the default content of “zzz_global_variables.R”:

#' A fix to avoid R CMD check notes for \"no visible binding for global variable\"
#'
#' This script makes it possible to refer to columns in data frame like objects
#' with Non Standard Evaluation (NSE) used by packages like data.table and dplyr
#' without encountering R CMD check notes concernng \"no visible binding for
#' global variable\". Names of the variables refered to with NSE are added to
#' a character vector in the function globalVariables() below.
#' 
#' @importFrom utils globalVariables
#' 
fix_undefined_global_vars <- function() {
  if (getRversion() >= "2.15.1")
    globalVariables(
      c(# Insert variable names below
        "foo_a", "foo_b"
        )
      )
}

fix_undefined_global_vars()

Add variables

It’s possible to maintain the character vector in globalVariables() manually. Below the column cyl from mtcars is added and the placeholders foo_a, foo_b are removed:

fix_undefined_global_vars <- function() {
  if (getRversion() >= "2.15.1")
    globalVariables(
      c(# Insert variable names below
        "cyl"
        )
      )
}

fix_undefined_global_vars()

However, this becomes cumbersome if NSE is used extensively. In stead,fixGlobals::add_global_vars() can be used to automatically add variable names to the vector.

For instance, a function subset_cars() can be defined to use data.table to subset several columns from mtcars:

subset_cars <- function(mtcars) {
  dtcars <- as.data.table(mtcars)
  dtcars[, .(cyl, mpg, disp, hp, drat, carb)]
}

In this example R CMD check notes will be provided for mpg, disp, hp, drat, carb if devtools::check() is executed. No note is provided for cyl since this variable was provided manually to globalVariables() earlier — see example above. However, if fixGlobals::add_global_vars() is executed zzz_global_variables.R will be updated and check notes concerning these variable are avoided:

fix_undefined_global_vars <- function() {
  if (getRversion() >= "2.15.1")
    globalVariables(
      c(# Insert variable names below
        "cyl", "carb", "disp", "drat", 
        "hp", "mpg"
        )
      )
}

fix_undefined_global_vars()

Notice that fixGlobals also keeps order by sorting alphabetically and puts no more that four variables in a line.

Install

You can install fixGlobals from GitHub.

# install.packages("devtools")
devtools::install_github("epetrovski/fixGlobals")