Incorrect replacement of preg_split by Nette\Utils\Strings::split() for regexp with subpattern

[rugabarbo]

Subject	Details
Rector version	dev-master@225f873
PHP version	PHP 7.3.11
Full Command	vendor/bin/rector process src tests --config rector-ci.yaml --dry-run

Current Behaviour

Source: https://travis-ci.org/cpliakas/git-wrapper/jobs/615485754#L322-L353

What is the problem

Nette Utils uses flag PREG_SPLIT_DELIM_CAPTURE by default when splitting lines:
https://github.com/nette/utils/blob/f1dc08536b040d5b569ee63354343145cbe6c591/src/Utils/Strings.php#L435

Because of this, Strings::split() also returns delimiters in the result set in cases where regexp contains subpatterns. Therefore, the Rector’s recommendation is incorrect, because in these cases method Strings::split() returns a different data set in the array.

[TomasVotruba]

What would be correct behavior and how to detect it?

[TomasVotruba]

Code to replicate:

$content = 'Hi my name is Tom';
$pattern = '#(\s)#'; // '#\s#' works fine, without "()"

$nativePieces = preg_split($pattern, $content);
var_dump($nativePieces);

$utilsPieces = \Nette\Utils\Strings::split($content, $pattern);
var_dump($utilsPieces);

// should be true
var_dump($nativePieces === $utilsPieces);   

[TomasVotruba]

This fixes it:

// @see https://stackoverflow.com/a/57787941/1348344
$utilsPieces = \Nette\Utils\Strings::split($content, $pattern, 0 & ~PREG_SPLIT_DELIM_CAPTURE);
dump($utilsPieces);