Remove %>% in favor of |>?

[billdenney]

We currently export %>%.

The built-in |> was added in R 4.1 (about 2 years ago). I think that it would make sense to remove the magrittr pipe in favor of the built-in pipe throughout the package.

[sfirke]

I was just thinking this myself as I reviewed #559 . I agree it's reasonable to stop using %>% in the code and to stop exporting it.

[olivroy]

The only thing is that the tidyverse has the policy of supporting the last 5 versions of R. I know that Hmisc depending on R 4.1 for this exact workflow broke CI for packages like ggplot2 on R 4.0 for example.

I know that purrr uses |> in examples. tidyverse/purrr#938 and has a hack for avoiding builds to work on 4.0 and 3.6.

I think it would be a good idea to stop re-exporting it as a first step? That would be a breaking change technically, but I am not sure many people rely on janitor to get the magrittr pipe in the first place.

[sfirke]

The stop re-exporting it seems like a good first step, it would be the breaking part for users. We should bump the janitor version number to indicate that. I think the next release could be 3.0.

[olivroy]

If the pipe is not re-exported, the examples using %>% will not work without an explicit library(magrittr) call. I may look into implementing the workaround used in tidyverse packages https://fosstodon.org/@gaborcsardi/111821576530612948

[billdenney]

Thanks for looking that up.

To me, that seems like a lot of effort for a relatively small gain. When R 4.0 is very old (for some now-uncertain definition of "very"), I think that we should switch to |> when the effort is near-zero.

I will close this for now.