Group_by then filters using dplyr

Question

First of all, please let me know if I am using dplyr badly because I am not sure if I will be the best. I have the following data framework:

mydf = data.frame(user = c(7,7,7,7,7,7,7,8,8,8,8,8,8),
                  col1 = c('0','0','1','1','0','3','NULL','3','3','0','1','0','0'),
                  col2 = runif(n=13),
                  col3 = letters[1:13], 
                  stringsAsFactors = FALSE)

> mydf
   user col1      col2 col3
1     7    0 0.7607907    a
2     7    0 0.1580448    b
3     7    1 0.8063540    c
4     7    1 0.7331512    d
5     7    0 0.2433631    e
6     7    3 0.2357065    f
7     7 NULL 0.4864172    g
8     8    3 0.6806089    h
9     8    3 0.2229874    i
10    8    0 0.6187911    j
11    8    1 0.7617177    k
12    8    0 0.5884821    l
13    8    0 0.4985750    m

The filtering I would like to do is a bit verbose, but I will try - I would like to filter out the dataframe by deleting all rows where col1 == '0' if this row occurs AFTER the first row for this user, where col1 == '1 ' . (bold means I messed up the original question and switched 0 and 1).

, 7, col1 == '1', 3, col1 == '0' ( 5). 8 11- , col1 == '1', 12 13, col1 == '0'.

:

> mydf
   user col1      col2 col3
1     7    0 0.7607907    a
2     7    0 0.1580448    b
3     7    1 0.8063540    c
4     7    1 0.7331512    d
6     7    3 0.2357065    f
7     7 NULL 0.4864172    g
8     8    3 0.6806089    h
9     8    3 0.2229874    i
10    8    0 0.6187911    j
11    8    1 0.7617177    k

, . rownums, , , , . , - :

mydf %>%
  mutate(rownums = 1:nrow(mydf)) %>%
  group_by(user) %>%
  filter(!(col1 == "0" & rownums > min(which(col1 == "1"))))

# A tibble: 9 x 5
# Groups:   col0 [2]
   user  col1      col2  col3 rownums
  <dbl> <chr>     <dbl> <chr>   <int>
1     7     0 0.2088034     a       1
2     7     0 0.2081894     b       2
3     7     1 0.1825428     c       3
4     7     1 0.2143353     d       4
5     7     3 0.1979774     f       6
6     7  NULL 0.2990799     g       7
7     8     3 0.7808038     h       8
8     8     3 0.1694272     i       9
9     8     1 0.1526450     k      11

, 10 .

!

EDIT - , group_by()% > % filter() - R dplyr. 99% group_by() summary(), , , .

EDIT2 - , !

mydf %>%
  group_by(col0) %>%
  mutate(rownums = 1:length(col0)) %>%
  filter(!(col1 == "0" & rownums > min(which(col1 == "1"))))

mutate() group_by() mutate(), , . .

Answer 1

cumany, , :

mydf %>%
  group_by(user) %>% 
  mutate(seen_one = cumany(col1 == "1")) %>%
  filter(!seen_one | col1 != "0")

"1" "" seen_one, , . ( filter "" , !(A & B) == !A | !B.)

Answer 2

dplyr

library(dplyr)

df %>% 
 group_by(user) %>% 
 mutate(id1 = row_number(), new_col = max(which(col1 == 1)+1)) %>% 
 filter(!(col1 == 0 & id1 >= new_col))

,

# A tibble: 10 x 6
# Groups:   user [2]
    user  col1       col2  col3   id1 new_col
   <dbl> <chr>      <dbl> <chr> <int>   <dbl>
 1     7     0 0.54742608     a     1       5
 2     7     0 0.89271859     b     2       5
 3     7     1 0.48999057     c     3       5
 4     7     1 0.17163211     d     4       5
 5     7     3 0.96146770     f     6       5
 6     7  NULL 0.31368382     g     7       5
 7     8     3 0.82051455     h     1       5
 8     8     3 0.30705440     i     2       5
 9     8     0 0.18545358     j     3       5
10     8     1 0.04834678     k     4       5

Answer 3

May be resolved by a small update to your attempt:

library(dplyr)
mydf %>%
   group_by(user) %>%
   filter(col1 != 0 | row_number() < which.max(col1 == 1))


#    user   col1        col2  col3
#   <dbl> <chr>       <dbl> <chr>
# 1     7     0 0.756522673     a
# 2     7     0 0.168314555     b
# 3     7     1 0.977254798     c
# 4     7     1 0.722721694     d
# 5     7     3 0.407849378     f
# 6     7  NULL 0.245335151     g
# 7     8     3 0.003423735     h
# 8     8     3 0.191716738     i
# 9     8     0 0.626846893     j
#10     8     1 0.546459621     k

Using filter, we select all rows where col10 is not equal or rows in which the current row is less than the index of the first occurrence 1 for this group.

Answer 4

in accordance with the @liborm answer directive:

mydf %>%
  group_by(user) %>%
  mutate(k = cumany(col1 == '0'), j = cumany(col1 == '1')) %>%
  filter(!(col1 == 0 & k == TRUE & j == TRUE)) %>%
  select(-k,-j)

returns:

    user  col1  col2  col3
   <dbl> <chr> <dbl> <chr>
 1     7     0     1     a
 2     7     0     1     b
 3     7     1     0     c
 4     7     1     0     d
 5     7     3     0     f
 6     7  NULL     1     g
 7     8     3     0     h
 8     8     3     1     i
 9     8     0     1     j
10     8     1     0     k