Cumsum with reset when 0 is encountered and by groups

Question

Below is my dataframe, I\'d like to get the \"yes\" column. I can\'t seem to get the cumsum to reset when it hits the 0 based on the \"value\" field by \"id\". Th

Answer 1

I think the answer by Imo could be an advertisement for the ta.table package (as if it needed yet another one for such a great package.) But I also think that base-R solutions should be attempted, so here's mine. It uses ave (which requires the FUN argument to be named) and does the cumsum operation twice, the first application is to create a grouping vector, and the second application creates the sequence. The second application could have also been seq.int but it would have seemed a bit clumsy since it would have needed to be function(x)seq.int(0,length(x)-1) since the default call to seq.int starts from 1 rather than 0.

test$yes2 <- ave(test$value, cumsum(test$value==0), FUN=cumsum)

> test
   id value yes yes2
1   1     1   1    1
2   1     1   2    2
3   1     0   0    0
4   1     1   1    1
5   2     1   1    2
6   2     1   2    3
7   2     1   3    4
8   2     1   4    5
9   3     0   0    0
10  3     1   1    1
11  3     1   2    2
12  3     0   0    0
13  4     1   1    1
14  4     1   2    2
15  4     0   0    0
16  4     0   0    0

Answer 2

You can create a new by variable on the fly like this:

test[, wrong := cumsum(value), by=.(id, tempID=cumsum(value==0))]
test
    id value correct wrong
 1:  1     1       1     1
 2:  1     1       2     2
 3:  1     0       0     0
 4:  1     1       1     1
 5:  2     1       1     1
 6:  2     1       2     2
 7:  2     1       3     3
 8:  2     1       4     4
 9:  3     0       0     0
10:  3     1       1     1
11:  3     1       2     2
12:  3     0       0     0
13:  4     1       1     1
14:  4     1       2     2
15:  4     0       0     0
16:  4     0       0     0

Note that test <- is not necessary here, as := will update the data.table by reference.