WISH: attr() allows to provide a default fallback value

[zeehio]

Wish

attr() allows to specify default=value to return a default value if the requested attribute is not found.

x <- 1

# current:
w <- attr(x, "hello")
if (is.null(w)) {
  w <- "world"
}

# wish:
w <- attr(x, "hello", default="world")

PS: big fan of your work. Thanks for this repository and for all your packages! 😃

[HenrikBengtsson]

I like this idea - this addition to attr() would be consistent to how getOption() and Sys.getenv() provide "default" fallbacks. Not sure if there are other functions too with this too.

FWIW, I wasn't sure whether NULL was a valid attribute value. It turns out it's not; it's not possible to have attributes will value NULL;

> x <- 42L; attributes(x) <- alist(abc = 1, def = NULL, ghi = 3)
> str(attributes(x))
List of 2
 $ abc: num 1
 $ ghi: num 3

In other words, one does not have to account for that as a special case and your "current" snippet is sufficient.

PS: big fan of your work. Thanks for this repository and for all your packages! smiley

Thanks. Happy to hear :)

[nfultz]

Might want to disable partial matching when a default is provided?

[vh-d]

I prefer more generic approach like this in my code

w <- ifnull(attr(x, "hello"), "world")

Would be nice to have something like this in base but the addition would potentially break someone's code of course.

[HenrikBengtsson]

Might want to disable partial matching when a default is provided?

Good point.

Though, I can see how "magically" setting exact = TRUE when default is specified, can become a point of confusion and possibly result in a blocking discussion for implementing support for a new argument default. If exact = TRUE becomes the new default, which I hope it one day will, this would be moot.

An alternative to having the default value of exact depend on whether argument default i specified, is to require exact = TRUE whenever default is specified. This means you'd had to write:

w <- attr(x, "hello", exact = TRUE, default = "world")

for now, to avoid an error. This would also help to prepare for the day when exact = TRUE becomes the new default (I hope).

(Just my thoughts)