Empty strings in result of "abc".split("")

[frewsxcv]

fn main() {
    println!("{:?}", ("abc".to_owned()).split("").collect::<Vec<_>>());
}

["", "a", "b", "c", ""]

Is this expected behavior? If so, it should probably be documented.

[frewsxcv]

I guess it's sort of in the same category as:

If a string contains multiple contiguous separators, you will end up with empty strings in the output:

and

This can lead to possibly surprising behavior when whitespace is used as the separator. This code is correct:

Which is in the documentation here.

[steveklabnik]

I am on my phone at the moment, but it is expected and documented IIRC.

On May 26, 2016, 12:39 -0400, Corey Farwellnotifications@github.com, wrote:

fnmain() { println!("{:?}",("abc".to_owned()).split("").collect::<Vec<_>>());}
["", "a", "b", "c", ""]

Is this expected behavior? If so, it should probably be documented.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly orview it on GitHub(#33882)

[est31]

The examples given in the documentation are

let x = "||||a||b|c".to_string();
let d: Vec<_> = x.split('|').collect();

assert_eq!(d, &["", "", "", "", "a", "", "b", "c"]);

and

let x = "    a  b c".to_string();
let d: Vec<_> = x.split(' ').collect();

assert_eq!(d, &["", "", "", "", "a", "", "b", "c"]);

The text doesn't go further than "will contain empty strings", but the examples suggest that each time the separator is directly followed by another occurence of it in the string, an empty string will be the output.

From reading that specific section of the documentation, I can't really find out the output of the split method for the empty string.

The empty string is a real edge case and at least deserves to be documented IMO, or, if it is considered invalid input, it should panic.

[frewsxcv]

The empty string is a real edge case and at least deserves to be documented IMO

I agree with this. Especially since doing split("") in other languages results in different behavior.

[frewsxcv]

JavaScript:

"abc".split("")
["a", "b", "c"]

Python:

>>> "abc".split("")
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: empty separator

Ruby:

irb(main):002:0> "abc".split("")
=> ["a", "b", "c"]

[est31]

C/C++:

#include <string.h>
#include <stdio.h>

int main(void)
{
    char input[] = "abc";
    char delim[] = "";
    char *token = strtok(input, delim);
    printf("[");
    while (token) {
        printf("\"%s\"", token);
        token = strtok(NULL, delim);
        if (token) {
            printf(", ");
        }
    }
    printf("]\n");
}

outputs

["abc"]

[nagisa]

The behaviour is expected (algorithm and behaviour we implement is greedy).

Consider "".split(""): in the string being split, there’s one empty string you can split on (greedily, that is), adjoined by (…at most…, remember, algorithm is greedy) two other empty strings, which is the result (a ["", ""]). Similar reasoning can be applied to the original report.

It should be documented if it isn’t already. (To me it looks like it isn’t)

[nagisa]

This seems like an easy enough to do change, thus marking as E-easy.

This is the documentation that needs to be edited, other examples for the function should be used as a reference. The algorithm behaviour is described in the comment above. In case anything is not clear, ask either here or on IRC (#rust-internals/#rust-libs).

[bluss]

Previous related discussion in #25986