pivot_wider not working with class <Period>

[bertrandh]

Brief description of the problem:
pivot_wider throws a strange error when used with a value_from column of type <Period> (as output from lubridate).
I am using R-3.6.3, lubridate_1.7.4, dbplyr_1.4.2.

Reprex:

df_0 <- tibble(period_name   = c("a", "b", "c"), 
               period_value  = c("1:1:1", 
                                 "2:2:2", 
                                 "4:4:4")) %>% 
         mutate(period_value = lubridate::hms(period_value))

df_0
## A tibble: 3 x 2
#  period_name period_value
#  <chr>       <Period>
#1 a           1H 1M 1S
#2 b           2H 2M 2S
#3 c           4H 4M 4S

Now, let's try widening the data frame

df_0 %>% pivot_wider(names_from = period_name, values_from = period_value)
#Error: No common type for `value` <Period> and `x` <Period>.

This error does not make sense to me.
BTW, at least it throws an error, whereas spread silently messed it up (see Stackoverflow question).

It works fine, though, with date_time type:

df_1 <- tibble(dttm_name   = c("a", "b", "c"), 
               dttm_value  = c("2020-01-01 1:1:1", 
                               "2020-02-03 2:2:2", 
                               "2020-05-10 4:4:4")) %>% 
         mutate(dttm_value = lubridate::as_datetime(dttm_value))

df_1
## A tibble: 3 x 2
#  dttm_name dttm_value
#  <chr>     <dttm>
#1 a         2020-01-01 01:01:01
#2 b         2020-02-03 02:02:02
#3 c         2020-05-10 04:04:04

df_1 %>% pivot_wider(names_from = dttm_name, values_from = dttm_value)
## A tibble: 1 x 3
#  a                   b                   c
#  <dttm>              <dttm>              <dttm>
#1 2020-01-01 01:01:01 2020-02-03 02:02:02 2020-05-10 04:04:04

Is that something wrong with pivot_wider (and spread) or with lubridate's Period class?

[hadley]

Duplicate of r-lib/vctrs#919