Chapter 8 R Language Characteristics
8.1 Help
Use ? to see the document of a function. One can even use ? to see the document of ?: try ?"?". The document, however, sometimes does not help a lot. The target audience of R document is usually not for beginners so may contain examples or wordings that are hard to understand. If one is consulting third-party packages not included in the core R, the quality of documents are indeed varying. Some authors provides good documentation while the others doesn’t. That said, documents are only one source of help for R learners. Usually a better way is by googling and searching on stackoverflow.
8.2 Vectorization
Every built-in data structure in R is of vector form. This means that many of the built-in functions that operate on these data structures actually operates on vectors. Consider the following minimal example:
1:10 + 1## [1] 2 3 4 5 6 7 8 9 10 11Yes the add (+) function is vectorized. An expereicend R user won’t write code like this:
res <- integer(10)
for ( i in 1:10 )
res[i] <- i + 1
res## [1] 2 3 4 5 6 7 8 9 10 11The performance of vectorized codes and non-vectorized one can be huge:
sillyAdd <- function(n) {
res <- integer(n)
for ( i in 1:n ) res[i] <- i + 1
res
}
system.time(res1 <- 1:1e6 + 1)## user system elapsed
## 0.044 0.005 0.051system.time(res2 <- sillyAdd(1e6))## user system elapsed
## 0.897 0.013 0.919identical(res1, res2)## [1] TRUEDue to performance (in most cases) gain, one should always consider using vectorization whenever possible.
One thing usually mis-understood by beginners: Are the apply family vectorized? The anwser is disppointedly a NO. apply family is a functional way (without side-effect) to write down codes intending to do loop-like tasks. They are NOT vectorization of any kind. This also means they do NOT gain any performance advantage.
8.3 Recycling
Recycling usually happens when an operation involves in at least two vectors, and one is shorter than the other. Take extraction by logical vector for example:
vec <- 1:10
vec[c(T, F)]## [1] 1 3 5 7 9The logical vector c(T, F) is recycled to make sure it is as long as the vector to be extracted. Recycling often occurs without awaring users. No warning. No message. Recyclingis everywhere:
1:10 + 1## [1] 2 3 4 5 6 7 8 9 10 11Here the shorter vector (which is 1) is recycled to have length 10. To see things more clear:
identical(1:10 + 1:2, 1:10 + rep(1:2, 5))## [1] TRUE8.4 The apply Family
The apply family is really one outstanding feature in the core of R language. Among the family, the most useful two are apply and lapply. Sometimes sapply also comes in handy.
8.4.1 apply
apply will call the given function column-by-column or row-by-row, according to the MARGIN argument supplied.
(mm <- matrix(1:12, 4, 3))## [,1] [,2] [,3]
## [1,] 1 5 9
## [2,] 2 6 10
## [3,] 3 7 11
## [4,] 4 8 12apply(mm, 1, sum) # by the 1st dimension: row## [1] 15 18 21 24apply(mm, 2, sum) # by the 2nd dimension: column## [1] 10 26 42apply use ellipsis, so it is very easy to apply functions with additional arguments:
mm <- matrix(1:12, 4, 3)
apply(mm, 2, mean, trim=.2) # apply the trimmed mean over each column## [1] 2.5 6.5 10.5apply(mm, 2, paste, collapse=':') # concatenate all numbers for each column## [1] "1:2:3:4" "5:6:7:8" "9:10:11:12"or lambdas:
mm <- matrix(1:12, 4, 3)
apply(mm, 2, function(x) sd(x) / mean(x))## [1] 0.5163978 0.1986145 0.1229519apply can also be used on data.frame in exactly the same way:
str(cars)## 'data.frame': 50 obs. of 2 variables:
## $ speed: num 4 4 7 7 8 9 10 10 10 11 ...
## $ dist : num 2 10 4 22 16 10 18 26 34 17 ...apply(cars, 2, max)## speed dist
## 25 1208.4.2 lapply and sapply
lapply is a list apply, so it operates on a list.
ll <- list(1:10, letters)
lapply(ll, length)## [[1]]
## [1] 10
##
## [[2]]
## [1] 26lapply always returns a list of the same length as its input. Sometimes it is sufficient to return only an atomic vector. Thgat’s where sapply comes in handy:
sapply(ll, length)## [1] 10 26Most of the time the call to sapply will have the effect of a unlist(lapply(...)).