testPipes

This package simply aims to reproduce an issue encountered while working on a bigger project when using pipes (%>%).

ISSUE

GHA fails to build on macOS and Ubuntu when pipes are used in more than one single file when pipes do not start on a new line. No errors locally (Linux).

IS THIS A BUG? NO

BRIEF SOLUTION TO THE ISSUE

On GHA, remember to install the package that you are testing by running R CMD INSTALL . on Linux and macOS systems.

Add this to your yaml file after 'Installing dependencies'

- name: Install package
if: runner.os == 'macOS' || runner.os == 'Linux'
run: R CMD INSTALL .

Detailed description of issue

examples of breaking code according to lintr

Example of code that breaks on GHA:

data.frame(col1 = c1, col2 = c2) %>% stats::na.omit()

Everything on one line: lintr does not like this!

On GHA, lintr::expect_lint_free() throws the following error:
Warning: no visible global function definition for ‘%>%’

examples of working code according to lintr

Example of code that builds without erros on GHA:

data.frame(col1 = c1,
          col2 = c2) %>% stats::na.omit()

Each pipe is on a new line: lintr likes this! <3

linter analysis

After some digging, it seems that the object_usage_linter is responsible for breaking the code. This linter harnesses the codetools package. This linter calls codetools::checkUsage on closures. In this specific case, this will look something like this:

codetools::checkUsage(function1 <- function(x) {
  c1 <- as.numeric(x$col1)
  c2 <- as.numeric(x$col2)
  data.frame(col1 = c1, col2 = c2) %>% stats::na.omit()
})