Allow @inheritDotParams to pull params from implementation (e.g. mutate.data.frame) like @inheritParams

[orgadish]

?dplyr::mutate

Testing using dev roxygen2@head 7a1dd39

@inheritParams is able to access the .by parameter, but @inheritDotParams is not (see examples below). Note: one workaround could be to specify @inheritDotParams dplyr:::mutate.data.frame but that didn't work either, with error:

✖ In topic 'baz2': @inheritDotParams failed to find topic "dplyr:::mutate.data.frame" in current package.
✖ In topic 'baz2': @inheritDotParams failed.
ℹ No arguments inherited from `dplyr:::mutate.data.frame()`.

The goal of this is to enable RStudio auto-completion of tidylog::mutate to dplyr::mutate.data.frame parameters. See elbersb/tidylog#77 for more information.

Reprex:

#' Foo
#'
#' @inheritParams dplyr::mutate
foo <- function(.data, .by = NULL, ...) { }

?foo

#' Bar
#'
#' @inheritDotParams dplyr::mutate
bar <- function(.data, ...) { }

?bar

#' Baz
#'
#' @inheritParams dplyr::mutate
#' @inheritDotParams dplyr::mutate
baz <- function(.data, ...) { }

?baz

[hadley]

As the message suggests, there's no topic called mutate.data.frame. You'll want to use dplyr::mutate instead.

[orgadish]

@hadley If you use @inheritDotParamas dplyr::mutate it doesn't pull any of the arguments from mutate.data.frame, so it only pulls .data

[hadley]

Ok, so the problem is this:

#' Foo
#'
#' @inheritDotParams dplyr::mutate
foo <- function(.data, .by = NULL, ...) {}

Apologies for missing this originally, but you included a lot of variations that did work making it harder for me to see what the problem is.

[orgadish]

@hadley Apologies for the confusion. In my original reprex I used both @inheritParams and @inheritDotParams and then I wanted to show that either together or separate it doesn't bring the desired params. But I ended up making the repro more confusing as a result.

The problem is

#' Bar
#'
#' @inheritDotParams dplyr::mutate
bar <- function(.data, ...) { }

Where we are hoping to use ... to pull .by and .keep, etc but it does not. Presumably because those exist only in the mutate.data.frame implementation which is not seen by @inheritDotParams.

[orgadish]

@hadley Take a look at the PR I've proposed in #1847 and see if that fits with what you had in mind.

[hadley]

I was simultaneously working on a different approach in #1848. I'd rather make @inheritParams and @inheritDotParams work in the same way, especially since I want to add the same filtering to @inheritParams. This will increase the chance of false positives, but I think it's just too confusing if we don't unify the behaviour of the two.

[orgadish]

Yes I agree that approach makes more sense!