Strange bahaviour with lapply, lazy evaluation? - r

Question

can someone explain me what is going on here?
I have a list of lists that I need to match with a table, and I am using lapply with fmatch (package fastmatch http://cran.r-project.org/web/packages/fastmatch/index.html) for that (which I think uses matching by hashing the table to be matched with, in contrast to match).
However, this is rather slow if table values have to be evaluated in the function (at least that's what I suspect), but I am not entirely sure.
I have found a workaround which speeds up the computation from 5.5 to 0.01s, but would like a more elegant solution.
Here is a reproducible example:
set.seed(10)
matchFeatures <- replicate(n = 1000, paste0("a", sample(x = 1:10000, size = sample(x = 1:10, size = 1))))
matchTable <- 1:10000
system.time(m1 <- lapply(matchFeatures, function(features) fmatch(features, paste0("a", 1:10000))))
system.time(m2 <- lapply(matchFeatures, function(features) force(fmatch(features, paste0("a", 1:10000)))))
system.time({tempTable <- paste0("a", 1:10000); m3 <- lapply(matchFeatures, function(features) fmatch(features, tempTable))})
identical(m1, m3)
Thanks Justin, just to follow up, I was looking for something like this:
system.time(m4 <- lapply(matchFeatures, fmatch, table = paste0("a", 1:10000)))

Answer 1

In the first two functions, you're running the paste command once for each iteration (i.e. 10000 times). In the third, it only happens once. If you use matchTable <- paste('a', 1:10000) and pass matchTable to all three versions you get a substantial speed up as expected.
matchFeatures <- replicate(n = 1000,
paste0("a",
sample(x = 1:10000,
size = sample(x = 1:10, size = 1))))
matchTable <- paste('a', 1:10000)
system.time(m1 <- lapply(matchFeatures,
function(features) fmatch(features, matchTable)))
system.time(m2 <- lapply(matchFeatures,
function(features) force(fmatch(features, matchTable))))
system.time(m3 <- lapply(matchFeatures,
function(features) fmatch(features, matchTable)))
identical(m1, m3)