How can I write a function that is iterable? - r

Question

I need to modify a function (below) that will apply row-wise with dplyr::mutate to remove any '_' characters and capitalise the first letter of each word.
My function
simple_cap <- function(x) {
s <- strsplit(x, "_")[[1]]
paste(toupper(substring(s, 1,1)), substring(s, 2),
sep="", collapse=" ")
}
My data
df <- read.table(text = c('
location obs
1 australia 12454.
2 new_south_wales 3931.
3 victoria 3244.
4 queensland 2477.
5 south_australia 834.
6 western_australia 1335.
7 tasmania 246.'), stringsAsFactors = F)
The dplyr::mutate call:
df %>% mutate(
location = simple_cap(location)
)
The output
location obs
1 Australia 12454
2 Australia 3931
3 Australia 3244
4 Australia 2477
5 Australia 834
6 Australia 1335
7 Australia 246
How can I change my function so that it can be used to iterate over the values in df$location rather than replacing them all with the output from the first element?

Answer 1

1) With gsub
We can use gsub to select the lower case characters ([a-z]), capture as a group ((...)) that is the first letter of the string (^) or (|) that follows an underscore (_) and replace with the backreference after converting to upper case (\\U)
Wrap with another gsub to remove the _ and replace with " "
df %>%
mutate(location = gsub("_", " ", gsub("(^|_)([a-z])", "\\1\\U\\2", location, perl = TRUE)))
# location obs
#1 Australia 12454
#2 New South Wales 3931
#3 Victoria 3244
#4 Queensland 2477
#5 South Australia 834
#6 Western Australia 1335
#7 Tasmania 246
2) With stringi
Or another option is stri_trans_totitle from stringi
library(stringi)
df %>%
mutate(location = stri_trans_totitle(stri_replace_all_fixed(location, "_", " ")))
# location obs
#1 Australia 12454
#2 New South Wales 3931
#3 Victoria 3244
#4 Queensland 2477
#5 South Australia 834
#6 Western Australia 1335
#7 Tasmania 246
3) Using OP's modified function
The strsplit output is a list. In the OP's code, it is just subsetting the first element by extracting [[1]]. But, here we have a list of length 7. So, one option is to use map from purrr (or with lapply/sapply from base R) and then do the pasteing of the substring
simple_cap <- function(x) {
s <- strsplit(x, "_")
purrr::map_chr(s, ~
paste(toupper(substring(.x, 1,1)), substring(.x, 2),
sep="", collapse=" "))
}
df %>%
mutate(location = simple_cap(location))
# location obs
#1 Australia 12454
#2 New South Wales 3931
#3 Victoria 3244
#4 Queensland 2477
#5 South Australia 834
#6 Western Australia 1335
#7 Tasmania 246
4) OP's modified function with sapply
simple_cap <- function(x) {
s <- strsplit(x, "_")
sapply(s, function(.s)
paste(toupper(substring(.s, 1,1)), substring(.s, 2),
sep="", collapse=" "))
}
5) No external packages
But, this can be done without using any external package
df$location <- gsub("_", " ", gsub("(^|_)([a-z])", "\\1\\U\\2", df$location, perl = TRUE))

Answer 2

There is a str_to_title function in stringr which capitalises the first character of word and with gsub we replace all the "_" (underscore) with " " (blank space).
library(stringr)
library(dplyr)
df %>%
mutate(location = str_to_title(gsub("_", " ", location)))
# location obs
#1 Australia 12454
#2 New South Wales 3931
#3 Victoria 3244
#4 Queensland 2477
#5 South Australia 834
#6 Western Australia 1335
#7 Tasmania 246

Answer 3

Ronak Shah and akrun have solved your specific problem. Here's the general solution to your title question (how do I write a function that is iterable).
In the parlance of R, you want a vectorized function -- a function that accepts a vector input and returns a vector output. There are two ways to do this.
1) Make sure each step in your function can accept a vector input and return a vector output. #akrun's 4th answer identifies the step in your code that prevents it from doing this, s <- strsplit(x, "_")[[1]].
2) Turn a non-vectorized function into a vectorized one with Vectorize. Option 1 is more efficient, but sometimes it's not possible. This is clearly an example where it's possible, but to show you how this works, lets vectorize your function with Vectorize
simple_cap <- function(x) {
s <- strsplit(x, "_")[[1]]
paste(toupper(substring(s, 1,1)), substring(s, 2),
sep="", collapse=" ")
}
simple_cap_v <- Vectorize(simple_cap, USE.NAMES = FALSE)
simple_cap(df$location)
# [1] "Australia"
simple_cap_v(df$location)
# [1] "Australia" "New South Wales" "Victoria" "Queensland"
# [5] "South Australia" "Western Australia" "Tasmania"
df %>% mutate(
location = simple_cap_v(location)
)
# location obs
# 1 Australia 12454
# 2 New South Wales 3931
# 3 Victoria 3244
# 4 Queensland 2477
# 5 South Australia 834
# 6 Western Australia 1335
# 7 Tasmania 246
Vectorize returns a function that is a wrapper to mapply. Effectively, a call to simple_cap_v(x) is now mapply(simple_cap, x, USE.NAMES = FALSE)