I've got a relatively simple problem (I think) and I want to solve it in a fast and efficient way.
I want to count the number of different elements in a vector up to each point in this vector.
For example, in a vector like this
vec <- c("a", "b", "c", "a", "a", "c", "d", "a")
I want to get the following vector of equal size as a result:
[1 2 3 3 3 3 4 4]
I could solve this of course with a for loop in combination with cumsum():
vec <- c("a", "b", "c", "a", "a", "c", "d", "a")
res <- T
for (i in 2:length(vec)) {
res[i] <- !(vec[i] %in% vec[1:(i-1)])
}
cumsum(res)
[1] 1 2 3 3 3 3 4 4
However, I am dealing with vectors that have several million elements and a for-loop approach takes forever for such a relatively simple problem.
I have the intuition that this should be solvable much faster and more clever. Do you have any ideas? Thank you!
(In case you're interested: I need this for a vocabulary growth curve analysis where we want to know at each point in the text how many different words, i.e. types, have been observed so far.)
Use cumsum on the non (!) duplicated values:
cumsum(!duplicated(vec))
#[1] 1 2 3 3 3 3 4 4
And another approach with match:
uni <- vector(length = length(vec))
uni[match(unique(vec), vec)] <- TRUE
cumsum(uni)