How to return the last value in a vector that met a certain condition

Question

I have a vector (in a data frame) filled with increasing numbers. I would like to find all consecutive numbers and replace them with the first number from the series. Is this possible to do without a loop?

My input data is:

V1
1
4
5
7
10
15
16
17
20

What I would like output is:

V1    Out
1     1
4     4
5     4
7     7
10    10
15    15
16    15
17    15
20    20

So far, I managed to calculate the difference between two rows using diff() and loop through the vector to replace the right values.

V1 <- c(1, 4, 5, 7, 10, 15, 16, 17, 20)
df <- data.frame(V1)
df$diff <- c(0, diff(df$V1) == 1)
df$Out <- NA
for (j in 1:(nrow(df))){
    if (df$diff[j] == 0){
        df$Out[j] <- df$V1[j]
    } else {
        df$Out[j] <- df$V1[max(which(df$diff[1:j] == 0))]
    }
}

It does the job, but it is very inefficient. Is there a way to get rid of the loop and make this code fast?

Thank you very much!

Answer 1

Using base R you can do,

with(d1, ave(V1, cumsum(c(1, diff(V1) != 1)), FUN = function(i) i[1]))
#[1]  1  4  4  7 10 15 15 15 20

dplyr

library(dplyr)

d1 %>% 
 group_by(grp = cumsum(c(1, diff(V1) != 1))) %>% 
 mutate(out = first(V1))

data.table

library(data.table)

setDT(d1)[, out := first(V1), by = cumsum(c(1, diff(V1) != 1))]

Answer 2

Another option, in 3 steps, using zoo package:

Define V2 as V1:

df$V2 <- df$V1

Replace the consecutive value (where diff is 1) by NA:

df$V2[c(FALSE, diff(df$V1)==1)] <- NA

Finally, use zoo::na.locf to replace NAs with last value:

library(zoo)
df$V2 <- na.locf(df$V2)

Output:

df
#   V1 V2
# 1  1  1
# 2  4  4
# 3  5  4
# 4  7  7
# 5 10 10
# 6 15 15
# 7 16 15
# 8 17 15
# 9 20 20

Another writting, in one line, using magrittr:

library(magrittr)
df$V2 <- df$V1 %>% replace(c(FALSE, diff(df$V1)==1), NA) %>% na.locf