Vectorization loop with repeated indexes
I have a vector of indices that contains duplicate values:
IN <- c(1, 1, 2, 2, 3, 4, 5)
I would like to use these indices to subtract two vectors:
ST <- c(0, 0, 0, 0, 0, 0, 0)
SB <- c(1, 1, 1, 1, 1, 1, 1)
However, I would like to do the subtraction in "order" so that after subtracting the first index values (0, 1), the second expression "works" with the first subtraction. I would like to get a vector FN that looks like this:
c(-2, -2, -1, -1, -1, 0, 0)
It's easy enough to do this in a for loop:
for(i in seq_along(IN)){
ST[IN[i]] <- ST[IN[i]] - SB[IN[i]]
}
But I need to run this loop many times on long vectors and it can take many hours. Is there a way to vectorize this task and avoid the for loop? Perhaps using the data.table technique?
Of course with data.table, this is
library(data.table)
DT = data.table(ST)
mDT = data.table(IN, SB)[, .(sub = sum(SB)), by=.(w = IN)]
DT[mDT$w, ST := ST - mDT$sub ]
ST
1: -2
2: -2
3: -1
4: -1
5: -1
6: 0
7: 0
Or with base R:
w = sort(unique(IN))
ST[w] <- ST[w] - tapply(SB, IN, FUN = sum)
# [1] -2 -2 -1 -1 -1 0 0
Here's an option using aggregate
R in base:
ag <- aggregate(.~IN, data.frame(IN, ST[IN]-SB[IN]), sum)
replace(ST, ag[,1], ag[,2])
#[1] -2 -2 -1 -1 -1 0 0
OR using xtabs
:
d <- as.data.frame(xtabs(B~A, data.frame(A=IN, B=ST[IN]-SB[IN])))
replace(ST, d[,1], d[,2])