[Rd] NextMethod() and argument laziness

[Tomas Kalibera]

On 8/7/19 3:10 PM, Davis Vaughan wrote:
> Hi all, I'd like to ask if the following behavior is a bug. To me it
> certainly feels surprising, at the very least. In this example, I would
> like to call NextMethod() from my `child` object, have `cols` be left
> untouched, and then substitute(cols) in the parent method. It works when
> you use a `parent` object (as expected), but I would have also expected to
> get `mpg` back when calling it from the `child` method.
>
> my_generic <- function(x, cols) {
>    UseMethod("my_generic")
> }
> my_generic.parent <- function(x, cols) {
>    substitute(cols)
> }
> my_generic.child <- function(x, cols) {
>    NextMethod()
> }
> obj_parent <- structure(mtcars, class = c("parent", class(mtcars)))
> obj_child <- structure(obj_parent, class = c("child", class(obj_parent)))
>
> my_generic(obj_parent, mpg)
> #> mpg
>
> my_generic(obj_child, mpg)
> #> cols

This works as documented, see ?NextMethod:

"the arguments will be the same in number, order and name as those to 
the current method but their values will be promises to evaluate their 
name in the current method and environment"

Hence NextMethod needs to create a promise with a name matching the 
formal argument name of the function from which NextMethod is invoked. 
Note this is not the same as with UseMethod, and this is also why 
substitute() works differently with UseMethod.

Best
Tomas

>
> Thanks,
> Davis
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel