add_count returns grouped dataframe

[davidchall]

According to the dplyr docs, add_count() is supposed to:

group transiently, so the output has the same groups as the input.

Although this is true for the dplyr version, I discovered that the dbplyr version does change the groups.

library(tidyverse)
library(dbplyr)

df <- tribble(
  ~name,    ~gender,   ~runs,
  "Max",    "male",       10,
  "Sandra", "female",      1,
  "Susan",  "female",      4
)

# ungrouped output
df %>% add_count(gender, wt = runs)
#> # A tibble: 3 x 4
#>   name   gender  runs     n
#>   <chr>  <chr>  <dbl> <dbl>
#> 1 Max    male      10    10
#> 2 Sandra female     1     5
#> 3 Susan  female     4     5

# grouped output
memdb_frame(df) %>% add_count(gender, wt = runs)
#> # Source:   lazy query [?? x 4]
#> # Database: sqlite 3.34.1 [:memory:]
#> # Groups:   gender
#>   name   gender  runs     n
#>   <chr>  <chr>  <dbl> <dbl>
#> 1 Sandra female     1     5
#> 2 Susan  female     4     5
#> 3 Max    male      10    10

^{Created on 2021-03-11 by the reprex package (v1.0.0)}

[davidchall]

I think this might be related to how dplyr_reconstruct handles dbplyr tables, in combination with #413.

[mgirlich]

Unfortunately, there needs to be done something in dplyr to fix this. See tidyverse/dplyr#5837

[davidchall]

Excellent - thanks for digging into this!

[davidchall]

Looks like this will be addressed by tidyverse/dplyr#5848