Learn R Programming

lintr (version 3.1.2)

nested_ifelse_linter: Block usage of nested ifelse() calls

Description

Calling ifelse() in nested calls is problematic for two main reasons:

  1. It can be hard to read -- mapping the code to the expected output for such code can be a messy task/require a lot of mental bandwidth, especially for code that nests more than once

  2. It is inefficient -- ifelse() can evaluate all of its arguments at both yes and no (see https://stackoverflow.com/q/16275149); this issue is exacerbated for nested calls

Usage

nested_ifelse_linter()

Arguments

Tags

efficiency, readability

Details

Users can instead rely on a more readable alternative modeled after SQL CASE WHEN statements.

Let's say this is our original code:

ifelse(
  x == "a",
  2L,
  ifelse(x == "b", 3L, 1L)
)

Here are a few ways to avoid nesting and make the code more readable:

  • Use data.table::fcase()

    data.table::fcase(
      x == "a", 2L,
      x == "b", 3L,
      default = 1L
    )
    

  • Use dplyr::case_match()

    dplyr::case_match(
      x,
      "a" ~ 2L,
      "b" ~ 3L,
      .default = 1L
    )
    

  • Use a look-up-and-merge approach (build a mapping table between values and outputs and merge this to the input)

    default <- 1L
    values <- data.frame(
      a = 2L,
      b = 3L
    )
    found_value <- values[[x]]
    ifelse(is.null(found_value), default, found_value)
    

See Also

linters for a complete list of linters available in lintr.

Examples

Run this code
# will produce lints
lint(
  text = 'ifelse(x == "a", 1L, ifelse(x == "b", 2L, 3L))',
  linters = nested_ifelse_linter()
)

# okay
lint(
  text = 'dplyr::case_when(x == "a" ~ 1L, x == "b" ~ 2L, TRUE ~ 3L)',
  linters = nested_ifelse_linter()
)

lint(
  text = 'data.table::fcase(x == "a", 1L, x == "b", 2L, default = 3L)',
  linters = nested_ifelse_linter()
)

Run the code above in your browser using DataLab