rows_update() is not sf friendly like other dplyr functions

[andresrcs]

When applied to an sf object, rows_update() throws the error message #> Error: Attempting to update missing rows. but if I transform the object to a tibble first, it works just fine, and then I can get back to an sf object with sf::st_as_sf(), which is not an ideal workflow. Could you make rows_update() function sf friendly like other dplyr functions?

library(dplyr)
library(rnaturalearth)

ne_states(country = "peru", returnclass = "sf") %>%
    select(name, name_len) %>% 
    filter(name %in% c("Arequipa", "Lima")) %>% 
    rows_update(tibble(name = "Arequipa", name_len = 0), by = "name")
#> Error: Attempting to update missing rows.

ne_states(country = "peru", returnclass = "sf") %>%
    select(name, name_len) %>% 
    filter(name %in% c("Arequipa", "Lima")) %>% 
    as_tibble() %>% 
    rows_update(tibble(name = "Arequipa", name_len = 0), by = "name")
#> # A tibble: 2 x 3
#>   name     name_len                                                     geometry
#>   <chr>       <int>                                           <MULTIPOLYGON [°]>
#> 1 Arequipa        0 (((-71.48848 -17.31799, -71.49771 -17.29713, -71.51452 -17.…
#> 2 Lima            4 (((-76.24919 -13.32044, -76.27428 -13.28729, -76.30695 -13.…

^{Created on 2021-06-04 by the reprex package (v2.0.0)}

[romainfrancois]

Can you have a look @krlmlr ?

[krlmlr]

Thanks. My first instinct is to suggest implementing rows_update() and friends in the sf package, even if it's with a wrapper as_tibble() %>% rows_update() %>% st_as_sf() . I'll take a look and propose something.

[jbeaulie]

This issue was 'closed as completed' but the problem documented in the original post persists with dplyr_1.1.4, except the error message has changed:
Error in rows_update():
! Can't reconstruct data frame.
✖ The [ method for class <sf/data.frame> must return a data frame with 1 column.
ℹ It returned a <sf/data.frame> of 2 columns.