Spliting string with a specific pattern - r

Question

I have a table with a character field that can have either of these pattern :
input
97 # a single number
210 foo # a number and a word
87 bar 89 # a number, a word, a number
21 23 # two numbers
123 2 fizzbuzz # two number, a word
12 fizz 34 buzz # a number, a word, a number, a word
I'd like to split each line up to 4 parts, containing respectively the first number, the first word if it exists, the second number if it exists, and the second word if it exists. So my example would give :
input nb_1 word_1 nb_2 word_2
97 97
210 foo 210 foo
87 bar 89 87 bar 89
21 23 21 23
123 2 fizzbuzz 123 2 fizzbuzz
12 fizz 34 buzz 12 fizz 34 buzz
Please note the case of two number, a word (the example before the last one) : it has nothing in word_1 as there is no word between the two numbers.
Is there a way to do this without a tedious if / if / else structure ?
If it can help, all the words belong to a list of 10 specific words. Also, if there are two words, they can be the same or different. Also, the numbers can be one, two or three digits long.
Thanks

Answer 1

Here is an idea using gsub and cSplit from splitstackshape package,
library(splitstackshape)
df$num <- gsub('\\D', ' ', df$V1)
df$wrds <- gsub('\\d', ' ', df$V1)
newdf <- cSplit(df, 2:3, ' ', 'wide')
newdf
# V1 num_1 num_2 wrds_1 wrds_2
#1: 97 97 NA NA NA
#2: 210 foo 210 NA foo NA
#3: 87 bar 89 87 89 bar NA
#4: 21 23 21 23 NA NA
#5: 123 2 fizzbuzz 123 2 fizzbuzz NA
#6: 12 fizz 34 buzz 12 34 fizz buzz
The only problem is row 5, which can be fixed as follows,
newdf$wrds_1 <- as.character(newdf$wrds_1)
newdf$wrds_2 <- as.character(newdf$wrds_2)
newdf$wrds_2[grep('[0-9]+\\s+[0-9]+\\s+[A-Za-z]', newdf$V1)] <- newdf$wrds_1[grep('[0-9]+\\s+[0-9]+\\s+[A-Za-z]', newdf$V1)]
newdf$wrds_1[grep('[0-9]+\\s+[0-9]+\\s+[A-Za-z]', newdf$V1)] <- NA
which finally gives,
newdf
# V1 num_1 num_2 wrds_1 wrds_2
#1: 97 97 NA NA NA
#2: 210 foo 210 NA foo NA
#3: 87 bar 89 87 89 bar NA
#4: 21 23 21 23 NA NA
#5: 123 2 fizzbuzz 123 2 NA fizzbuzz
#6: 12 fizz 34 buzz 12 34 fizz buzz
DATA
dput(df)
structure(list(V1 = c("97", " 210 foo", " 87 bar 89",
" 21 23", " 123 2 fizzbuzz",
" 12 fizz 34 buzz")), .Names = "V1", row.names = c(NA,
-6L), class = "data.frame")

Answer 2

Tried in a different way...
library(splitstackshape)
abc <- data.frame(a=c(97,"210 foo","87 bar 89","21 23","123 2 fizzbuzz","12 fizz 34 buzz"))
abc1 <- data.frame(cSplit(abc, "a", " ", stripWhite = FALSE))
abc <- cbind(abc,abc1)
names(abc) <- c("input","nb_1", "word_1", "nb_2","word_2")
abc[,1:5] <-apply(abc[,1:5] , 2, as.character)
for(i in 1:nrow(abc)){
abc$word_2[i] <- replace(abc$word_2[i] , is.na(abc$word_2[i]),abc$nb_2[grepl("[a-z]",abc$nb_2[i])][i])
abc$nb_2[i] <- replace(abc$nb_2[i] , is.na(abc$nb_2[i])|grepl("[a-z]",abc$nb_2[i]),abc$word_1[grepl("[0-9]",abc$word_1[i])][i])
}
abc$word_1 <- ifelse(grepl("[0-9]",abc$word_1),NA,abc$word_1)
abc[is.na(abc)] <- ""
print(abc)
input nb_1 word_1 nb_2 word_2
1 97 97
2 210 foo 210 foo
3 87 bar 89 87 bar 89
4 21 23 21 23
5 123 2 fizzbuzz 123 2 fizzbuzz
6 12 fizz 34 buzz 12 fizz 34 buzz

Answer 3

This is a hacky function to do it... although you might have other cases that would break it.
f <- function(x){
string2 <- strsplit(x, " ")[[1]]
if (length(string2) < 2)
return(c(string2, NA, NA, NA))
arenums <- grepl("\\d", string2)
c(string2[which(arenums)[1]],
if (arenums[2]) NA else string2[which(!arenums)[1]],
string2[which(arenums)[2]],
if (arenums[2]) string2[which(!arenums)[1]] else string2[which(!arenums)[2]])
}
> f("97")
[1] "97" NA NA NA
> f("210 foo")
[1] "210" "foo" NA NA
> f("87 bar 89")
[1] "87" "bar" "89" NA
> f("21 23")
[1] "21" NA "23" NA
> f("123 2 fizzbuzz")
[1] "123" NA "2" "fizzbuzz"
> f("12 fizz 34 buzz")
[1] "12" "fizz" "34" "buzz"