complete(fill) should accept single value

[hadley]

library(tidyr)

df <- tibble(x = c(1, NA), y = c(1, NA))
df |> complete(fill = 0)
#> Error in `replace_na()`:
#> ! `replace` must be a list.

^{Created on 2022-04-13 by the reprex package (v2.0.1)}

This would make it consistent with pivot_wider(values_fill) and friends.

[DavisVaughan]

We seem to go back and forth on how we feel about this 😄 #1108 (comment) and #1108 (comment)

Implementing this would mean that every column in df would have to be the same type (because the 1 replace value would be applied to every column), which seems extremely unlikely for the typical analysis.

It would also make code harder to read. i.e. is replace_na(x, 1) replacing values in a vector x or a data frame x?

[hadley]

Oh yeah, you'll probably need to remind me about this again in a couple of years 😄