Multiple for loop time computation very high in R - r

Question

I have data about machines in the following form
Number of rows - 900k
Data
A B C D E F G H I J K L M N
---- -- --- ---- --- --- --- --- --- --- --- --- --- ---
1 1 1 1 1 1 1 1 1 1 0 1 1 0 0
2 0 0 0 0 1 1 1 0 1 1 0 0 1 0
3 0 0 0 0 0 0 0 1 1 1 1 1 0 0
1 indicates that the machine was active and 0 indicates that it was inactive.
I want my output to look like
A B C D E F G H I J K L M N
---- -- --- ---- --- --- --- --- --- --- --- --- --- ---
1 1 1 1 1 1 1 1 1 1 1 1 1 0 0
2 0 0 0 0 1 1 1 1 1 1 0 0 1 0
3 0 0 0 0 0 0 0 1 1 1 1 1 0 0
Basically all I am trying to do is look for zeros in a particular row and if that zero is surrounded by ones on either side, replace 0 with 1
example -
in row 1 you have zero in column J
but you also have 1 in column I and K
which means I replace that 0 by 1 because it is surrounded by 1s
The code I am using is this
for(j in 2:13) {
if(data[i,j]==0 && data[i,j-1]==1 && data[i,j+1]==1){
data[i,j] = 1
}
}
}
Is there a way to reduce the time computation for this? This takes me almost 30 mins to run in R. Any help would be appreciated.

Answer 1

this is faster because it does not require to iterate through the rows.
for(j in 2:13) {
data[,j] = ifelse(data[,j-1] * data[,j+1]==1,1,data[,j])
}
or a littlebit more optimized, without using ifelse
for(j in 2:(ncol(data) - 1)) {
data[data[, j - 1] * data[, j + 1] == 1, j] <- 1
}

Answer 2

You could also use gsub to replace any instances of 101 with 111 using the following code:
collapsed <- gsub('101', '111', apply(df1, 1, paste, collapse = ''))
data <- as_tibble(t(matrix(unlist(sapply(collapsed, strsplit, split = '')), nrow = numLetters)))
names(data) <- LETTERS[1:numLetters]
Here's a comparison of everyone's solutions:
library(data.table)
library(rbenchmark)
library(tidyverse)
set.seed(1)
numLetters <- 13
df <- as_tibble(matrix(round(runif(numLetters * 100)), ncol = numLetters))
names(df) <- LETTERS[1:numLetters]
benchmark(
'gsub' = {
data <- df
collapsed <- gsub('101', '111', apply(data, 1, paste, collapse = ''))
data <- as_tibble(t(matrix(unlist(sapply(collapsed, strsplit, split = '')), nrow = numLetters)))
names(data) <- LETTERS[1:numLetters]
},
'for_orig' = {
data <- df
for(i in 1:nrow(data)) {
for(j in 2:(ncol(data) - 1)) {
if(data[i, j] == 0 && data[i, j - 1] == 1 && data[i, j + 1] == 1) {
data[i, j] = 1
}
}
}
},
'for_norows' = {
data <- df
for(j in 2:(ncol(data) - 1)) {
data[, j] = ifelse(data[, j - 1] * data[, j + 1] == 1, 1, data[, j])
}
},
'vectorize' = {
data <- df
for(i in seq(ncol(data) - 2) + 1) {
condition <- data[, i - 1] == data[, i + 1] & data[, i - 1] == 1 & data[, i] == 0
data[which(condition), i] <- 1
}
},
'index' = {
data <- df
idx <- apply(data, 1, function(x) c(0, diff(x)))
data[which(idx == -1 & lead(idx == 1), arr.ind = TRUE)[, 2:1]] <- 1
},
replications = 100
)
The indexing solution (which has since been deleted) wins hands-down in terms of computational time for a 13-by-100 data frame.
test replications elapsed relative user.self sys.self user.child
3 for_norows 100 1.19 7.438 1.19 0 NA
2 for_orig 100 9.29 58.063 9.27 0 NA
1 gsub 100 0.28 1.750 0.28 0 NA
5 index 100 0.16 1.000 0.16 0 NA
4 vectorize 100 0.87 5.438 0.87 0 NA
sys.child
3 NA
2 NA
1 NA
5 NA
4 NA

Answer 3

Cut the time by using vectorized operations. As you are planning to do the same thing for every row, this can be done by utilizing the vectorized conditional statements.
for(i in seq(ncol(data) - 2) + 1){ #<== all but last and first column
#Find all neighbouring columns that are equal, where the the center column is equal to 0
condition <- data[, i - 1] == data[, i + 1] & data[, i - 1] == 1 & data[, i] == 0
#Overwrite only the values that holds the condition
data[which(condition), i] <- 1
}

Answer 4

You can avoid loops altogether and use indexing to replace all the values at once:
nc <- ncol(df)
df[, 2:(nc - 1)][df[, 1:(nc - 2)] * df[, 3:nc] == 1] <- 1