New linter encouraging %||%

[MichaelChirico]

This is now a base function (in r-devel) and so should be encouraging more broadly

r-devel/r-svn@aa23547

I would wait for it to make a release first, but leaving here as a reminder. Even if it's not eventually chosen for inclusion in a released version of R, we could offer such a linter for pointing to the {rlang} implementation.

[MichaelChirico]

Wrote this ad hoc linter to help find usages in r-devel:

is_null_linter <- make_linter_from_xpath(lint_message = "%||%", "
  //SYMBOL_FUNCTION_CALL[text() = 'is.null']
    /parent::expr
    /parent::expr[
      (
        preceding-sibling::IF
        and (
          expr[2] = following-sibling::ELSE/following-sibling::expr
          or expr[2][LEFT_ASSIGN]/expr[1] = following-sibling::ELSE/following-sibling::expr
        )
      ) or (
        preceding-sibling::OP-EXCLAMATION
        and parent::expr/preceding-sibling::IF
        and (
          expr[2] = parent::expr/following-sibling::expr[1]
          or expr[2][LEFT_ASSIGN]/expr[1] = parent::expr/following-sibling::expr[1]
        )
      )
    ]
")

r-devel/r-svn@5cf7f69

[AshesITR]

Sounds good. We should maybe see if some %||% definitions "in the wild" are exported and test for more than is.null.
rlang::`%||%` is fine, ours is different but not exported.

https://github.com/search?q=lang%3AR+%2F%60%25%5C%7C%5C%7C%25%60+%3C-%2F&type=code

Turns up with some alternative definitions.
Here are a few:

• https://github.com/ivan-rivera/RedditExtractor/blob/6962176962aa39f1bba4134d919a91810d05b4a7/R/helpers.R#L31
• https://github.com/wlandau/instantiate/blob/f910e354d4f8635b93df589f1b9cde3df86eedda/R/utils_logic.R#L1
• https://github.com/reconhub/reconhub.github.io/blob/af9f6f2f4ead6a7eb4ac65b25ab759bb83884851/renv/activate.R#L62
• https://github.com/sckott/apipkgen/blob/acd1f2a7fb73c6407daf54e39bccb9c245fa0762/R/zzz.R#L37
• https://github.com/carpentries/lesson-transition/blob/573c40bff1847644b07c26670e51080dedfb2470/create-success-comment.R#L15

[MichaelChirico]

As I see it, once %||% is part of base R (i.e. in a released version), assuming that definition is safe. We assume base exports are not overwritten in many places already.

[AshesITR]

I agree. Should we rename our version ahead of time?

[MichaelChirico]

Yes... %|||%? 😅

Might also be better to find places where we truly need the length(x) == 0 || is.na(x) parts and use ours, using the simple base version otherwise.

[IndrajeetPatil]

What are we going to call this linter?

Should it be null_coalescing_linter() (to be consistent with other languages; e.g., JavaScript, C#, etc. that also have this operator)?

[MichaelChirico]

I worry not many folks will make the cross-language connection (FWIW the docs officially cite Ruby as the inspiration https://chrisarcand.com/null-coalescing-operators-and-rubys-conditional-assignments/), and the name is pretty long.

I'd lean towards coalesce_operator, the downside is it collides with dplyr::coalesce(). We might consider just linting both things related to %||% and things related to dplyr::coalesce and data.table::fcoalesce under the same linter (somewhat analogous to conjunct_test_linter()).