How to drop unused labels from a memisc:data.set in R? - r

Question

I want to drop all unsued labels from a data.set.
Let's assume this example data.set (which is class from the memisc package).
library(memisc)
d <- data.set(a = sample(1:10), b=rep(c(14,72),5))
labels(d$b) <- c('First' = 14, 'no-use' = 33, 'Second' = 72)
The resulting data.set:
Data set with 10 observations and 2 variables
a b
1 4 First
2 1 Second
3 9 First
4 8 Second
5 7 First
6 10 Second
7 5 First
8 3 Second
9 2 First
10 6 Second
You see that for b only two values used but it has three labels.
> labels(d$b)
Values and labels:
14 'First'
33 'no-use'
72 'Second'
How can I drop the unused label (33) from there? The point is all unsued labels should be droped and I don't know which one is unused. I would know how to remove 33 explicite. But that is not the goal.
I know from the basic-R data.frame the function droplevels(). Would be nice to have something like droplabels().

Answer 1

This isn't very compact, but you could use the following
labels(d$b) <- labels(d$b)[seq_len(length(unique(d$b)))]
update
Your question states you want to drop '72' when it looks like you want to drop '33'. Regardless, the following function will drop any unused labels
labels(d$b) <- labels(d$b)[labels(d$b)#values %in% unique(d$b)]
The following will drop all unused labels for all elements of a list
for (i in seq_along(d)) {
if(!is.null(labels(d[[i]]))) {
labels(d[[i]]) <- labels(d[[i]])[labels(d[[i]])#values %in% unique(d[[i]])]
}
}