str_split must return empty array for empty string

[mvorisek]

Description

The following code:

<?php

print_r(str_split('', 1));
print_r(mb_str_split('', 1));
print_r(mb_str_split('', 1, 'UTF-8'));
print_r(mb_str_split('', 1, 'ASCII'));
print_r(array_chunk([], 1));

https://3v4l.org/1WsGi

Resulted in but I expected this output instead:

 Array
 (
-    [0] => 
 )
 Array
 (
 )
 Array
 (
 )
 Array
 (
 )
 Array
 (
 )

I belive this is an inconsistency that should be fixed.

The inconsistency was introduced in 0818fae by a mistake as found out by @emkey08 in the discussion below.

PHP Version

PHP 7.4 - master

Operating System

any

[emkey08]

Wouldn't it actually be the correct solution instead if both str_split('') and mb_str_split('') would return an empty array?

Splitting a string is supposed to return an array of the string's characters. An empty string has zero characters, so the expected and logical result would be an empty array (that is, an array containing zero characters).

When splitting a string, it should hold that count(str_split($str)) is equivalent to strlen($str). However, this currently isn't the case if the input string is empty.

[emkey08]

When str_split was initially added to PHP 5.0, str_split('') actually returned an empty array. See here.

This commit changed it, but this obviously was an unintended (and also undocumented) side effect. The commit claims to be a performance optimization only, while in fact it was a functional change: the commit author simply didn't correctly consider the case when the input string is empty. All later PHP versions built upon this change, and so the current str_split behavior is actually based on a mistake.

It would of course be a breaking change to fix str_split('') to return an empty array, but note that the current implementation also causes things to break. It's quite common to use foreach (str_split($str) as $char) to iterate over a string, but you must not assume strlen($char) === 1 inside that loop, or your code will break. Bummer.

[emkey08]

As a final note, the current documentation also claims

The default length is 1, meaning every chunk will be one byte in size.

which obviously isn't true.

[mvorisek]

When str_split was initially added to PHP 5.0, str_split('') actually returned an empty array. See here.

This commit changed it, but this obviously was an unintended (and also undocumented) side effect. The commit claims to be a performance optimization only, while in fact it was a functional change: the commit author simply didn't correctly consider the case when the input string is empty. All later PHP versions built upon this change, and so the current str_split behavior is actually based on a mistake.

👍, that "optimization commit" should be fixed. Then it would be consistent also with array_chunk behaviour - https://3v4l.org/1QLKU .

[cmb69]

I tend to agree that this is a bug (although one that should not be fixed in a revision), and not merely a documentation issue.