Splitting a string into fixed-size chunks - r

Question

This seems like a really simple task, but I can't find a good solution in base R. I have a character string with 2N characters. How do I split this into a character vector of length N, with each element being a 2-character string?
I could use something like substr with Vectorize:
vss <- Vectorize(substr, c("start", "stop"))
ch <- paste(rep("a", 1e6), collapse="")
vss(ch, seq(1, nchar(ch), by=2), seq(2, nchar(ch), by=2))
but this is really slow for long strings (O(N^2) I believe).

Answer 1

If you want speed, Rcpp is always a good choice:
library(Rcpp);
cppFunction('
List strsplitN(std::vector<std::string> v, int N ) {
if (N < 1) throw std::invalid_argument("N must be >= 1.");
List res(v.size());
for (int i = 0; i < v.size(); ++i) {
int num = v[i].size()/N + (v[i].size()%N == 0 ? 0 : 1);
std::vector<std::string> resCur(num,std::string(N,0));
for (int j = 0; j < num; ++j) resCur[j].assign(v[i].substr(j*N,N));
res[i] = resCur;
}
return res;
}
');
ch <- paste(rep('a',1e6),collapse='');
system.time({ res <- strsplitN(ch,2L); });
## user system elapsed
## 0.109 0.015 0.121
head(res[[1L]]); tail(res[[1L]]);
## [1] "aa" "aa" "aa" "aa" "aa" "aa"
## [1] "aa" "aa" "aa" "aa" "aa" "aa"
length(res[[1L]]);
## [1] 500000
Useful reference: http://gallery.rcpp.org/articles/strings_with_rcpp/.
More demos:
strsplitN(c('abcd','efgh'),2L);
## [[1]]
## [1] "ab" "cd"
##
## [[2]]
## [1] "ef" "gh"
##
strsplitN(c('abcd','efgh'),3L);
## [[1]]
## [1] "abc" "d"
##
## [[2]]
## [1] "efg" "h"
##
strsplitN(c('abcd','efgh'),1L);
## [[1]]
## [1] "a" "b" "c" "d"
##
## [[2]]
## [1] "e" "f" "g" "h"
##
strsplitN(c('abcd','efgh'),5L);
## [[1]]
## [1] "abcd"
##
## [[2]]
## [1] "efgh"
##
strsplitN(character(),5L);
## list()
strsplitN(c('abcd','efgh'),0L);
## Error: N must be >= 1.
There are two important caveats with the above implementation:
1: It doesn't handle NAs correctly. Rcpp seems to stringify to 'NA' when it's forced to come up with a std::string. You can easily solve this in Rland with a wrapper that replaces the offending list components with a true NA.
x <- c('a',NA); strsplitN(x,1L);
## [[1]]
## [1] "a"
##
## [[2]]
## [1] "N" "A"
##
x <- c('a',NA); ifelse(is.na(x),NA,strsplitN(x,1L));
## [[1]]
## [1] "a"
##
## [[2]]
## [1] NA
##
2: It doesn't handle multibyte characters correctly. This is a tougher problem, and would require a rewrite of the core function implementation to use a Unicode-aware traversal. Fixing this problem would also incur a significant performance penalty, since you wouldn't be able to preallocate each vector in one shot prior to the assignment loop.
strsplitN('aΩ',1L);
## [[1]]
## [1] "a" "\xce" "\xa9"
##
strsplit('aΩ','');
## [[1]]
## [1] "a" "Ω"
##