Split vector of strings and paste subset of resulting elements into a new vector

Question

Define

z<- as.character(c(\"1_xx xx xxx_xxxx_12_sep.xls\",\"2_xx xx xxx_xxxx_15_aug.xls\"))

such that

> z
[1] \"1_xx xx x         


        

Answer 1

One call to gsub (and some regex magic based on @Andrie's answer) can do this. See ?regexp for details on what I used in the pattern and replacement (back-reference) arguments.

gsub("^(\\d+_).*_(\\d+_\\w*).xls", "\\1\\2", z)
# [1] "1_12_sep" "2_15_aug"

Answer 2

Using a bit of magic in the stringr package: I separately extract the left and right date fields, combine them, and finally remove the .xls at the end.

library(stringr)
l <- str_extract(z, "\\d+_")
r <- str_extract(z, "\\d+_\\w*\\.xls")
gsub(".xls", "", paste(l, r, sep=""))

[1] "1_12_sep" "2_15_aug"

str_extract is a wrapper around some of the base R functions which I find easier to use.

Edit Here is a short explanation of what the regex does:

• \\d+ looks for one or more digits. It is escaped to distinguish from a normal character d.
• \\w* looks for zero or more alphanumeric characters (word). Again, it's escaped.
• \\. looks for a decimal point. This needs to be escaped because otherwise the decimal point means any single character.

In theory the regex should be quite flexible. It should find single or double characters for your dates.

Answer 3

You can do this using a combination of strsplit, substr and lapply:

y <- strsplit(z,"_",fixed=TRUE)
lapply(y,FUN=function(x){paste(x[1],x[4],substr(x[5],1,3),sep="_")})

Answer 4

An alternative along the same lines of @Joran's Answer is this:

foo <- function(x) {
    o <- paste(x[c(1,4,5)], collapse = "_")
    substr(o, 1, nchar(o) - 4) 
}

sapply(strsplit(z, "_"), foo)

The differences are minor - I use collapse = "_" and nchar() but other than that it is similar.

You can write this as a one-liner

sapply(strsplit(z, "_"), 
       function(x) {o <- paste(x[c(1,4,5)], 
                               collapse = "_"); substr(o, 1, nchar(o)-4)})

but writing the custom function to apply is nicer.