如何用特殊方式填写NA

Question

I have 3 columns 2 of them are groups one of them is NA or a or b or c.

For each element of group1 in group2 I want fill NA with number as follow:

the first row of each element of group1 in group2 is a, it should start from 1 until to get another letter a or b or c. after that I want to add 1 in a row after a, b or c.

example:

     group1     group2       act
        1         1           a
        1         1           NA
        1         1           NA
        1         1            b
        1         1           Na
        1         1            a
        1         1            NA
        1         1            a
        1         2            a
        1         2            NA
        1         2             a
        2         1            a
        2         1            NA
        2         1            b
        2          1           b
        2          1           NA
        2         1            a

     group1     group2       act          New
        1         1           a            1
        1         1           NA           1
        1         1           NA           1
        1         1            b           1
        1         1           Na           2
        1         1            a           2
        1         1            NA          3
        1         1            a           3
        1         2            a           1
        1         2            NA          1
        1         2             a          1
        2         1            a           1
        2         1            NA          1
        2         1            b           1
        2          1           b           2
        2          1           NA          3
        2         1            a           3

if the pattern is not clear ask me to explain more

Answer 1

On the grouped data you can increment by 1 on each non-NA value by using cumsum() . The data is shifted down 1 row using lag() .:

library(dplyr)

df %>% 
  group_by(group1, group2) %>%
  mutate(new = lag(cumsum(!is.na(act)), default = 1))

# A tibble: 17 x 4
# Groups:   group1, group2 [3]
   group1 group2 act     new
    <int>  <int> <chr> <dbl>
 1      1      1 a         1
 2      1      1 NA        1
 3      1      1 NA        1
 4      1      1 b         1
 5      1      1 NA        2
 6      1      1 a         2
 7      1      1 NA        3
 8      1      1 a         3
 9      1      2 a         1
10      1      2 NA        1
11      1      2 a         1
12      2      1 a         1
13      2      1 NA        1
14      2      1 b         1
15      2      1 b         2
16      2      1 NA        3
17      2      1 a         3