Last week markbulling over at Drunks & Lampposts posted a method of using
sapply to filter a list by a predicate. Today the @RLangTip tip of the day was to use
sapply similarly. This made makes me wonder if R‘s very useful higher-order functions aren’t as well known as they should be. In this case, the
Filter higher-order function would be the tool to use.
Filter works more or less like the
*apply family of functions, but it performs the subsetting (the filtering) of a list based on a predicate in a single step.
As an example, let’s say we have a list of 1000 vectors, each of length 2 with \(x_1,\,x_2 \in [0,\,1]\), and we want to select only those vectors where the elements of the list sum to a value greater than 1. With
Filter, this is all we have to do:
mylist <- lapply(1:1000, function(i) c(runif(1), runif(1))) method.1 <- Filter(function(x) sum(x) > 1, mylist)
Which is at least a bit more transparent than the
method.2 <- mylist[sapply(mylist, function(x) sum(x) > 1)]
In some very quick tests, I found no performance difference between the two approaches.
There are other useful higher-order functions. If you are interested, check out