Force && conditions to be written separately where appropriate

For readability of test outputs, testing only one thing per call to testthat::expect_true() is preferable, i.e., expect_true(A); expect_true(B) is better than expect_true(A && B), and expect_false(A); expect_false(B) is better than expect_false(A || B).

Usage

conjunct_test_linter(
  allow_named_stopifnot = TRUE,
  allow_filter = c("never", "not_dplyr", "always")
)

Arguments

allow_named_stopifnot

Logical, TRUE by default. If FALSE, "named" calls to stopifnot(), available since R 4.0.0 to provide helpful messages for test failures, are also linted.

allow_filter

Character naming the method for linting calls to filter(). The default, "never", means filter() and dplyr::filter() calls are linted; "not_dplyr" means only dplyr::filter() calls are linted; and "always" means no calls to filter() are linted. Calls like stats::filter() are never linted.

Details

Similar reasoning applies to && usage inside stopifnot() and assertthat::assert_that() calls.

Relatedly, dplyr::filter(DF, A & B) is the same as dplyr::filter(DF, A, B), but the latter will be more readable / easier to format for long conditions. Note that this linter assumes usages of filter() are dplyr::filter(); if you're using another function named filter(), e.g. stats::filter(), please namespace-qualify it to avoid false positives. You can omit linting filter() expressions altogether via allow_filter = TRUE.

See also

linters for a complete list of linters available in lintr.

Tags

best_practices, configurable, package_development, pkg_testthat, readability

Examples

# will produce lints
lint(
  text = "expect_true(x && y)",
  linters = conjunct_test_linter()
)
#> ::warning file=<text>,line=1,col=1::file=<text>,line=1,col=1,[conjunct_test_linter] Instead of expect_true(A && B), write multiple expectations like expect_true(A) and expect_true(B) The latter will produce better error messages in the case of failure.

lint(
  text = "expect_false(x || (y && z))",
  linters = conjunct_test_linter()
)
#> ::warning file=<text>,line=1,col=1::file=<text>,line=1,col=1,[conjunct_test_linter] Instead of expect_false(A || B), write multiple expectations like expect_false(A) and expect_false(B) The latter will produce better error messages in the case of failure.

lint(
  text = "stopifnot('x must be a logical scalar' = length(x) == 1 && is.logical(x) && !is.na(x))",
  linters = conjunct_test_linter(allow_named_stopifnot = FALSE)
)
#> ::warning file=<text>,line=1,col=1::file=<text>,line=1,col=1,[conjunct_test_linter] Instead of stopifnot(A && B), write multiple conditions like stopifnot(A, B). The latter will produce better error messages in the case of failure.

lint(
  text = "dplyr::filter(mtcars, mpg > 20 & vs == 0)",
  linters = conjunct_test_linter()
)
#> ::warning file=<text>,line=1,col=23::file=<text>,line=1,col=23,[conjunct_test_linter] Use dplyr::filter(DF, A, B) instead of dplyr::filter(DF, A & B).

lint(
  text = "filter(mtcars, mpg > 20 & vs == 0)",
  linters = conjunct_test_linter()
)
#> ::warning file=<text>,line=1,col=16::file=<text>,line=1,col=16,[conjunct_test_linter] Use dplyr::filter(DF, A, B) instead of dplyr::filter(DF, A & B).

# okay
lint(
  text = "expect_true(x || (y && z))",
  linters = conjunct_test_linter()
)

lint(
  text = 'stopifnot("x must be a logical scalar" = length(x) == 1 && is.logical(x) && !is.na(x))',
  linters = conjunct_test_linter(allow_named_stopifnot = TRUE)
)

lint(
  text = "dplyr::filter(mtcars, mpg > 20 & vs == 0)",
  linters = conjunct_test_linter(allow_filter = "always")
)

lint(
  text = "filter(mtcars, mpg > 20 & vs == 0)",
  linters = conjunct_test_linter(allow_filter = "not_dplyr")
)

lint(
  text = "stats::filter(mtcars$cyl, mtcars$mpg > 20 & mtcars$vs == 0)",
  linters = conjunct_test_linter()
)