Looping through a column in R

Question

I am using the stats package R and would like to loop through column[x] in all the rows of a dataframe, operate on the data in each cell in the column with a function and pass the result to a new column (with the calculated result in the new column aligned with the data in column[x])

Two problems. 1) I can't get it to work and 2) looping seems to be discouraged in the R articles I've read. Is there an alternative approach and if not, does anyone have an example of how to carry out the loop? thanks.

Answer 1

Looping is not necessarily discouraged.

Get it working first, and only then think about getting if faster.

Answer 2

Without any examples, it's hard to know how to respond. The basic case of what you're describing, however, is this:

#Just a very simple data frame
dat <- data.frame(x = c(1, 2, 3))
#Compute the squared value of each value in x
dat$y <- dat$x^2
#See the resultant data.frame, now with column y
dat

When you tell R to square a vector (or vector-like structure, like dat$x), it knows to square each value separately. You don't need to explicitly loop over those values most of the time - although, as Dirk notes, you should only worry about optimizing your loops if they are causing you problems. That said, I certainly prefer reading and writing

dat$y <- dat$x^2

to:

for(i in 1:length(dat$x)){
  dat$y[i] <- dat$x[i]^2
}

... where possible.

Answer 3

if parse.smiles() is a function you want to apply to all the entry of a vector "vec", then you can use:

lapply(1:length(vec),parse.smiles(vec[i]))

Answer 4

The only reason looping is discouraged is that it is slow. R is designed to work on vectors at a time and has lots of functions to accomplish this. The whole apply family, as well as functions like Vectorize to help out. So the idiom is that if your using for loops you're not thinking in R, but sometimes for loops really are just appropriate.

To do this in the R way of thinking, Vectorize your function, if it is not already vectorized (see the Vectorize function) then call that function with the entire column as an argument and assign that to the new column.

f<-Vectorize(function(x,...),'x')
data$newcolumn<-f(data[,1])

The apply family (apply, sapply, lapply, mapply, tapply) are also alternatives. Most native R functions are already vectorized, but be careful when passing extra arguments that are supposed to be interpreted as vectors.