combine(NULL) should return logical() instead of error

[zeehio]

Given that combine() acts like c() I don't understand why combine() (without args) does not work like c() returning NULL and raises an error instead.

If I am combining a list that may contain all NULL elements, I'd rather get back a NULL than tryCatching or checking for that condition myself in advance. It's not a big deal, but it does not seem consistent with for instance bind_rows() that returns an empty tibble.

If you agreed I'd be happy to provide a pull request.

c(1, NULL, 2)
#> [1] 1 2
dplyr::combine(1, NULL, 2)
#> [1] 1 2
c()
#> NULL
dplyr::combine(NULL)
#> Error in combine_all(args): no data to combine, all elements are NULL
c(NULL, NULL)
#> NULL
dplyr::combine(NULL, NULL)
#> Error in combine_all(args): no data to combine, all elements are NULL

As always, thanks for your time

[krlmlr]

@romainfrancois: Do you remember why you made combine() raise an error in response to #1596? I feel this and combine(NULL) or combine(list()) should return logical(), for consistency with tibble::enframe(NULL) .

[romainfrancois]

I don’t recall. Why logical ?

[krlmlr]

Because logical() can be coerced to anything, see also tidyverse/tibble#352.

[romainfrancois]

That makes sense. Not sure why we went for strict error.

[krlmlr]

Thanks.

@zeehio: Would you like to submit a PR?

[zeehio]

I will be happy to submit a PR next week, if you don't mind the delay 😃

[zeehio]

@krlmlr submitted, feel free to review whenever you have the time 😄 #3387

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/