Benchmarking performance of row selection via `-which(…)` vs. `!…`.

bm.R

> library(microbenchmark)
> m <- data.frame(a = sample.int(10, 100000, replace=T),
                  b = sample.int(10, 100000, replace=T))
> microbenchmark(m[-which(m[, 1] == 4 & m[, 2] == 5), ], m[! (m[, 1] == 4 & m[, 2] == 5), ])
Unit: milliseconds
                                   expr      min       lq   median       uq      max neval
 m[-which(m[, 1] == 4 & m[, 2] == 5), ] 27.59480 29.64628 30.09091 30.69976 42.04620   100
      m[!(m[, 1] == 4 & m[, 2] == 5), ] 29.00876 29.70103 30.41829 33.10700 42.29671   100
> m <- as.matrix(m)
microbenchmark(m[-which(m[, 1] == 4 & m[, 2] == 5), ], m[! (m[, 1] == 4 & m[, 2] == 5), ])
Unit: milliseconds
                                   expr      min       lq   median       uq      max neval
 m[-which(m[, 1] == 4 & m[, 2] == 5), ] 9.826798 10.73116 11.15770 13.74899 19.66254   100
      m[!(m[, 1] == 4 & m[, 2] == 5), ] 9.290829 10.49009 11.08083 12.46361 18.83631   100