new_group_by.md

library(dplyr, warn.conflicts = FALSE)

df <- tibble(
  f1 = factor(c("a", "a", "a", "b", "b"), levels = c("a", "b", "c")), 
  f2 = factor(c("d", "e", "d", "e", "f"), levels = c("d", "e", "f")), 
  x  = c(1, 1, 1, 2, 2), 
  y  = 1:5
)


df %>%
  group_by(f1, f2) %>%
  filter(y > 3)
#> # A tibble: 2 x 4
#> # Groups:   f1, f2 [9]
#>   f1    f2        x     y
#>   <fct> <fct> <dbl> <int>
#> 1 b     e         2     4
#> 2 b     f         2     5

df %>%
  group_by(f1, f2) %>%
  filter(y > 3) %>%
  summarise(n = n())
#> # A tibble: 9 x 3
#> # Groups:   f1 [3]
#>   f1    f2        n
#>   <fct> <fct> <int>
#> 1 a     d         0
#> 2 a     e         0
#> 3 a     f         0
#> 4 b     d         0
#> 5 b     e         1
#> 6 b     f         1
#> 7 c     d         0
#> 8 c     e         0
#> 9 c     f         0

df %>%
  group_by(x) %>%
  filter(y > 3)
#> # A tibble: 2 x 4
#> # Groups:   x [2]
#>   f1    f2        x     y
#>   <fct> <fct> <dbl> <int>
#> 1 b     e         2     4
#> 2 b     f         2     5

df %>%
  group_by(x) %>%
  filter(y > 3) %>%
  summarise(n = n())
#> # A tibble: 2 x 2
#>       x     n
#>   <dbl> <int>
#> 1     1     0
#> 2     2     2

g <- df %>%
  group_by(x)

# corrupt the grouping structure
attributes(g)$groups <- attributes(g)$groups[1, ]

# the result is bad
g %>%
  summarise(n = n())
#> # A tibble: 1 x 2
#>       x     n
#>   <dbl> <int>
#> 1     1     3

g %>%
  mutate(n = n())
#> # A tibble: 5 x 5
#> # Groups:   x [2]
#>   f1    f2        x     y        n
#>   <fct> <fct> <dbl> <int>    <int>
#> 1 a     d         1     1        3
#> 2 a     e         1     2        3
#> 3 a     d         1     3        3
#> 4 b     e         2     4        0
#> 5 b     f         2     5 73211632

^{Created on 2018-12-11 by the reprex package (v0.2.1)}