Adding values sequentially since a zero in R

Question

Hi I'm trying to add up values in a vector since a zero value in R. I have a vector like this:

wdnew
 [1] 1 0 0 1 1 0 0 1 1 1 1 1

And I want to create a new vector that counts the number of iterations or places since a zero value. If I did this properly I'd get a vector like this:

wdseq
 [1] 1 0 0 1 2 0 0 1 2 3 4 5

This ought to be straightforward but I can't figure out how to generate the sequence, any suggestions? Thanks,

It doesn't matter for this time, but next time please remember to use `dput` to share your example data. — Hack-R, Aug 19 '16 at 23:05
You have people answering who are assuming there are only zeros and 1s in the vector. Please be more specific. Are there always only zeros and 1s? Or can the nonzero values be any number? — Rich Scriven, Aug 19 '16 at 23:16
Lots of answers, @Rose. Can you please "accept" your preferred answer (and potentially upvote any and all you find particularly good)? — r2evans, Aug 20 '16 at 00:10
See [this identical post](http://stackoverflow.com/questions/32501902/cumulative-sum-that-resets-when-0-is-encountered) — alexis_laz, Aug 20 '16 at 08:52

score 5 · Answer 1 · edited Aug 19 '16 at 23:16

5

One option is to use rleid to create a group variable and do cumsum by group:

wdnew <- c(1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1)
ave(wdnew, data.table::rleid(wdnew != 0), FUN=cumsum)
# [1] 1 0 0 1 2 0 0 1 2 3 4 5

edited Aug 19 '16 at 23:16

Hack-R

22,422
14
75
131

answered Aug 19 '16 at 22:43

Psidom

209,562
33
339
356

1

I think the `!= 0` is not necessary.. – Arun Aug 19 '16 at 23:53
2

@Arun It's not very clear if the original vector contains only 1 and 0. This is to handle that kind of case as pointed out by another user. – Psidom Aug 20 '16 at 00:03

r2evans · Answer 2 · 2016-08-19T23:26:22.757

4

An option without data.table:

wdnew <- c(1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1)
rl <- rle(wdnew)
unlist(mapply(function(a,b) b * seq_len(a), rl$lengths, rl$values))
#  [1] 1 0 0 1 2 0 0 1 2 3 4 5

As @DirtySockSniffer suggested, this can be combined into a single line:

with(rle(wdnew), unlist(mapply(function(a,b) b * seq_len(a), lengths, values)))

Or stepping off of @user20650's suggestion:

wdnew * sequence(rle(wdnew)$lengths)

I think this last one may win the codegolf ...

(All assuming that wdnew is all 0s and 1s.)

edited Aug 19 '16 at 23:26

answered Aug 19 '16 at 22:53

r2evans

141,215
6
77
149

not yet, though there are many advocates and many good reasons to use it, it just hasn't reached my "learn-priority" yet – r2evans Aug 19 '16 at 23:09
I would recommend the simplest answer that doesn't rely on external packages as the solution. Looks like this one. This one or the one by @ZheyuanLi – Hack-R Aug 19 '16 at 23:13

cgage1 · Answer 3 · 2016-08-19T23:09:13.187

3

If you want to use a loop 'cause you like using loops, you can do this:

nums = c(1,0,0,1,1,0,0,1,1,1,1,1)

runningCount = 0
for(i in 1:length(nums)){
   if(nums[i] == 1){ 
      runningCount = runningCount + 1
      nums[i] = runningCount
   } else {
      runningCount = 0
   }
}

nums
 [1] 1 0 0 1 2 0 0 1 2 3 4 5

edited Aug 19 '16 at 23:09

answered Aug 19 '16 at 22:48

cgage1

579
5
15

Zheyuan Li · Answer 4 · 2016-08-20T00:07:41.920

I'm trying to add up values in a vector since a zero

Maybe

k <- rle(as.logical(x))$lengths
ave(x, rep.int(1:length(k), k), FUN = cumsum)

Note, I am taking a more general setting, where your x is not just 0 /1 binary. But as.logical(x) will give a TRUE / FALSE binary result, indicating chunks of zeros (FALSE) and non-zeros (TRUE). If your x is surely 0 / 1 binary, just using

k <- rle(x)$lengths
ave(x, rep.int(1:length(k), k), FUN = cumsum)

is legitimate.

Example

set.seed(0); x <- sample(0:3, 20, TRUE)
# [1] 3 1 1 2 3 0 3 3 2 2 0 0 0 2 1 3 1 2 3 1

k <- rle(as.logical(x))$lengths    ## chunk size
# [1] 5 1 4 3 7

ave(x, rep.int(1:length(k), k), FUN = cumsum)    ## cumulative sum each chunk
# [1]  3  4  5  7 10  0  3  6  8 10  0  0  0  2  3  6  7  9 12 13

Hack-R · Answer 5 · 2016-08-19T23:29:03.923

3

I may win the award for longest answer on this one.

Well, at least I didn't call any packages and stuck to basic functions (FWIW).

vec <-  c(1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1); res  <- vec

for(i in 2:length(vec)){m <- 1
  while(vec[i-m] > 0 & vec[i]>0){m <- m+1}
      res[i] <- m; if(vec[i]==0) res[i] <- 0}

[1] 1 0 0 1 2 0 0 1 2 3 4 5

edited Aug 19 '16 at 23:29

answered Aug 19 '16 at 23:02

Hack-R

22,422
14
75
131

score 1 · Answer 6 · answered Aug 20 '16 at 04:22

1

Here is another approach with diff and ave

ave(wdnew, cumsum(c(TRUE, diff(wdnew == 1)!=0)), FUN = seq) * wdnew
#[1] 1 0 0 1 2 0 0 1 2 3 4 5

data

wdnew <- c(1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1)

answered Aug 20 '16 at 04:22

akrun

874,273
37
540
662

Adding values sequentially since a zero in R

6 Answers6

data