Relax 1d atomic vectors or lists requirement

[hadley]

As it's breaking many revdeps so will likely have deleterious impact on user code

[hadley]

@lionel- can you take on this one? I'm not going to have time since I'm teaching Thursday and Friday.

The problem arises from the use of as_tibble() in three places. On the whole, I think the increased strictness of tibble is a good thing but it causes problems with some base R functions which 1d arrays like (like scale(), as in #2374). I think the simplest solution probably to switch to tibble::new_tibble(), but it requires a little thought. Any solution should include a PR that ensures this works:

df <- data.frame(x = 1:10)
df$y <- scale(1:10)
  
p <- ggplot(df, aes(x, y)) + geom_point()
layer_data(p)

Beware that data.frame(x = scale(1:10)) strips the dim attribute that causes the problem.

It'd be great if you could check a couple of the revdeps that fail to make sure this actually fixes the problem (easiest way is to search https://docs.google.com/spreadsheets/d/1UZIFwt3j3XjnRGOMNJZvpal_CAj7Hz8aNXJt41m3JjM/edit#gid=0 for "atomic vector")

[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/