| whichVector.logical {R.utils} | R Documentation |
Identifies TRUE elements in a logical vector.
This method is faster than which() for logical vectors,
especially when there are no missing values.
## S3 method for class 'logical': whichVector(x, na.rm=TRUE, use.names=TRUE, ...)
x |
A logical vector of length N. |
na.rm |
If TRUE, missing values are treated as FALSE,
otherwise they are returned as NA. |
use.names |
If TRUE, the names attribute is preserved,
otherwise it is not return. |
... |
Not used. |
Returns an integer vector of length less or equal to N.
Simple comparison on R v2.7.1 on Windows XP, show that
this implementation can be more than twice as fast as
which(), especially when there are no missing
value (and na.rm=FALSE) is used.
Henrik Bengtsson (http://www.braju.com/R/)
which()
# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Simulate two large named logical vectors,
# one with missing values one with out
# - - - - - - - - - - - - - - - - - - - - - - - - - -
N <- 1e6;
# Vector #1
x <- sample(c(TRUE, FALSE), size=N, replace=TRUE);
names(x) <- seq_along(x);
# Vector #2
y <- x
y[sample(N, size=0.1*N)] <- NA;
# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Validate consistency
# - - - - - - - - - - - - - - - - - - - - - - - - - -
stopifnot(identical(which(x), whichVector(x)));
stopifnot(identical(which(y), whichVector(y)));
# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Benchmarking
# - - - - - - - - - - - - - - - - - - - - - - - - - -
# Number of iterations
K <- 5;
t1 <- 0;
for (kk in 1:K) {
t1 <- t1 + system.time({ idxs1 <- which(x) });
};
t2 <- 0;
for (kk in 1:K) {
t2 <- t2 + system.time({ idxs2 <- whichVector(x, na.rm=FALSE) });
};
cat(sprintf("whichVector(x, na.rm=FALSE)/which(x): %.2f\n", (t2/t1)[3]));
stopifnot(identical(idxs1, idxs2));
t1 <- 0;
for (kk in 1:K) {
t1 <- t1 + system.time({ idxs1 <- which(y) });
};
t2 <- 0;
for (kk in 1:K) {
t2 <- t2 + system.time({ idxs2 <- whichVector(y) });
};
cat(sprintf("whichVector(y)/which(y): %.2f\n", (t2/t1)[3]));
stopifnot(identical(idxs1, idxs2));