Faceting by POSIXct fails with uninformative error if timezone is UTC

[hadley]

library(ggplot2)
library(tibble)

df <- tibble(y = as.POSIXct("2018-01-01 00:00:00", tz = "UTC"), x = 110)
df %>% 
  ggplot(aes(x)) +
  geom_histogram() +
  facet_wrap(~y)
#> Error: `scale_id` must not be `NA`

df <- tibble(y = as.POSIXct("2018-01-01 00:00:00"), x = 110)
df %>% 
  ggplot(aes(x)) +
  geom_histogram() +
  facet_wrap(~y)
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

^{Created on 2021-02-05 by the reprex package (v1.0.0)}

[yutannihilation]

Thanks, this is a duplicate of #4175. This is because rbind_dfs() doesn't support timezone (see #4175 (comment)).

I think we have two points here:

1. facet should handle POSIXct with timezone properly.
2. the error message should be more informative (e.g. facets fail with uninformative error for some column names #4138 is a different issue, but shows the same message)

For point 1, we can keep discussing on #4175. By the way, do you have any thoughts on vctrs integration?

[hadley]

Ok, lets close this issue, and discuss in #4175. I did a very quick vctrs experiment in #4342; I think it's probably worth committing to for the next big ggplot2 release.

[yutannihilation]

Oh, the PR looks promising.

it's probably worth committing to for the next big ggplot2 release.

Agreed!

[thomasp85]

Full support of a switch to vctrs