Count the number of unique characters in a string - r

Question

I have a dataframe where one of the columns is of type string.
I would like to count the number of unique/distinct characters in that string.
eg.
"banana" -> 3
'he' -> 2
A reproducible example:
I have a data frame where a column is type string. I would need to filter out those rows where the string has only one distinct character.
col1 col2 col3
new york
qqqq
melbourne
aaaaaa
I would need to have a final data frame like
col1 col2 col3
new york
melbourne
So delete those rows completely.

Answer 1

This makes no assumption about "characters" being in letters and avoids making R data structures:
library(inline)
.char_unique_code <- "
std::vector < std::string > s = as< std::vector < std::string > >(x);
unsigned int input_size = s.size();
std::vector < std::string > chrs(input_size);
for (unsigned int i=0; i<input_size; i++) {
std::string t = s[i];
for (std::string::iterator chr=t.begin();
chr != t.end(); ++chr) {
if (chrs[i].find(*chr) == std::string::npos) {
chrs[i] += *chr;
}
}
}
return(wrap(chrs));
"
char_unique <-
rcpp(sig=signature(x="std::vector < std::string >"),
body=.char_unique_code,
includes=c("#include <string>",
"#include <iostream>"))
nchar(char_unique("banana"))
## [1] 3
Why avoid making R lists?
library(stringr)
library(microbenchmark)
library(ggplot2)
str_char_ct_unique <- function(x) sum(!!str_count(x, letters))
char_ct_unique <- function(x) nchar(char_unique(x))
r_char_ct_unique <- function(x) length(unique(strsplit(x, "")[[1]]))
microbenchmark(stringr=str_char_ct_unique("banana"),
rcpp=char_ct_unique("banana"),
r=r_char_ct_unique("banana"),
times=1000) -> mb
## Unit: microseconds
## expr min lq mean median uq max neval cld
## stringr 125.978 129.1765 139.271061 130.9415 139.3870 334.563 1000 c
## rcpp 1.458 2.0160 3.002184 2.6345 3.1365 32.244 1000 a
## r 4.797 6.1070 8.292847 7.3380 8.0505 86.709 1000 b
Let's make a vectorized version of Cath's pure R solution (not bothering with the other one since it's way too constrained) and compare against a vector of small random strings:
library(random)
library(purrr)
char_ct_unique <- function(x) nchar(char_unique(x))
r_char_ct_unique <- function(x) map_int(map(x, function(x) unique(strsplit(x, "")[[1]])), length)
tst <- as.vector(randomStrings(n=100, len=20, unique=FALSE))
sum(char_ct_unique(tst) == r_char_ct_unique(tst))
## [1] 100
microbenchmark(rcpp=char_ct_unique(tst),
r=r_char_ct_unique(tst),
times=1000)
## Unit: microseconds
## expr min lq mean median uq max neval cld
## rcpp 53.643 56.2375 66.69311 60.2740 68.178 250.992 1000 a
## r 683.420 759.4070 952.14407 822.8905 922.710 6513.508 1000 b
And, now for the 10,000 character random string:
dat <- readLines("https://gist.githubusercontent.com/hrbrmstr/f80b157b383134b37fb3/raw/534b4c79e7c51710c6db6961bc5dc5ec25c4242b/gistfile1.txt")
digest::digest(dat, "sha1", serialize=FALSE)
## [1] "6c6695dd2f314762c81e6e6891ec1c138a4f3a08"
nchar(dat)
## [1] 10000
char_ct_unique(dat) == r_char_ct_unique(dat)
## [1] TRUE
microbenchmark(rcpp=char_ct_unique(dat),
r=r_char_ct_unique(dat),
times=1000)
## Unit: microseconds
## expr min lq mean median uq max neval cld
## rcpp 73.801 110.681 122.9091 118.330 139.373 308.602 1000 a
## r 377.556 430.703 533.9120 448.631 492.466 4275.568 1000 b
I forgot to do David's "fixed" version:
f_r_char_ct_unique <- function(x) map_int(map(x, function(x) unique(strsplit(x, "", fixed=TRUE)[[1]])), length)
and, let's make it more interesting:
dat <- c(dat, toupper(dat), tolower(dat))
microbenchmark(rcpp=char_ct_unique(dat),
r=r_char_ct_unique(dat),
fr=f_r_char_ct_unique(dat),
times=1000)
## Unit: microseconds
## expr min lq mean median uq max neval
## rcpp 218.299 284.143 331.319 332.281 358.1215 696.907 1000
## r 1266.976 1442.460 1720.320 1494.167 1634.7870 5896.685 1000
## fr 1260.027 1444.298 1769.664 1501.416 1652.8895 78457.729 1000

Answer 2

We can use str_count
library(stringr)
sum(!!str_count(str1, letters))
#[1] 3
Update
Using the new dataset
i1 <- !sapply(df1$col1, function(x) any(str_count(x, letters)>1))
df1[i1,,drop=FALSE]
data
str1 <- "banana"