Perform an operation with temporary groups

This is an experimental new function that allows you to modify the grouping variables for a single operation.

with_groups(.data, .groups, .f, ...)

Arguments

.data	A data frame
.groups	<tidy-select> One or more variables to group by. Unlike group_by(), you can only group by existing variables, and you can use tidy-select syntax like c(x, y, z) to select multiple variables. Use NULL to temporarily ungroup.
.f	Function to apply to regrouped data. Supports purrr-style ~ syntax
...	Additional arguments passed on to ....

Examples

df <- tibble(g = c(1, 1, 2, 2, 3), x = runif(5))
df %>%
  with_groups(g, mutate, x_mean = mean(x))
#> # A tibble: 5 x 3
#>       g     x x_mean
#>   <dbl> <dbl>  <dbl>
#> 1     1 0.795  0.784
#> 2     1 0.774  0.784
#> 3     2 0.923  0.704
#> 4     2 0.485  0.704
#> 5     3 0.500  0.500
df %>%
  with_groups(g, ~ mutate(.x, x1 = first(x)))
#> # A tibble: 5 x 3
#>       g     x    x1
#>   <dbl> <dbl> <dbl>
#> 1     1 0.795 0.795
#> 2     1 0.774 0.795
#> 3     2 0.923 0.923
#> 4     2 0.485 0.923
#> 5     3 0.500 0.500

df %>%
  group_by(g) %>%
  with_groups(NULL, mutate, x_mean = mean(x))
#> # A tibble: 5 x 3
#> # Groups:   g [3]
#>       g     x x_mean
#>   <dbl> <dbl>  <dbl>
#> 1     1 0.795  0.695
#> 2     1 0.774  0.695
#> 3     2 0.923  0.695
#> 4     2 0.485  0.695
#> 5     3 0.500  0.695

# NB: grouping can't be restored if you remove the grouping variables
df %>%
  group_by(g) %>%
  with_groups(NULL, mutate, g = NULL)
#> # A tibble: 5 x 1
#>       x
#>   <dbl>
#> 1 0.795
#> 2 0.774
#> 3 0.923
#> 4 0.485
#> 5 0.500