Special considerations around base pipe `|>` and its placeholder `_`

[jennybc]

The introduction of the _ placeholder for the base pipe |> is likely to result in lots of reports of reprex() not capturing folks reprex or error.

Because invalid use of _ is going to result in a parse error, which is not something reprex can currently capture, because such code cannot be rmarkdown::render()ed.

@jthomasmock just shared these concrete examples with me.

In the Console, with R 4.2, one sees:

mtcars$disp |> mean
#> Error: The pipe operator requires a function call as RHS

or

mtcars$disp |> mean(_)
#> Error: pipe placeholder can only be used as a named argument

But an attempt to reprex() the expression mtcars$disp |> mean(_) results in:

ℹ Rendering reprex...
Error in parse(text = x, keep.source = TRUE) : 
  pipe placeholder can only be used as a named argument

This is really not a new phenomenon, i.e. we've long had users reporting similar situations where knitr's handling of parsing errors means the code can't be reprex()ed (e.g. #308, #359, #396). But I suspect increased use of |> means more people will bump up against this.