I think this is not easily vectorizable…. at least i do not know how.
You can do it by hand
via:
my_cumsum <- function(x){
grp = integer(length(x))
grp[1] = 1
for(i in 2:length(x)){
if(x[i-1] + x[i] <= 10){
grp[i] = grp[i-1]
x[i] = x[i-1] + x[i]
} else {
grp[i] = grp[i-1] + 1
}
}
data.frame(grp, x)
}
For your data this gives:
> my_cumsum(df$value)
grp x
1 1 4
2 1 9
3 2 7
4 2 10
5 3 8
6 3 9
7 4 2
8 4 7
9 4 10
10 5 6
11 5 8
12 6 6
13 6 9
14 6 10
15 7 4
Also for my “counter-example” this gives:
> my_cumsum(c(10,6,4))
grp x
1 1 10
2 2 6
3 2 10
As @Khashaa pointed out this can be implementet more efficiently via Rcpp
. He linked to this answer How to speed up or vectorize a for loop? which i find very useful