regex - how to split text into sentences without breaking brackets

nzhou

Hi,

I need a regex to split text into an array of sentences. For example, the below is the text:

$text = "aaa bbb (ccc ddd (eee fff) ggg (xxx yyy zzz)) hhh iii jjj (kkk lll mmm) nnn ooo ppp";

I want to split it into an array like this:

[0] => "aaa bbb"
[1] => "(ccc ddd (eee fff) ggg (xxx yyy zzz))"
[2] => "hhh iii jjj"
[3] => "(kkk lll mmm)"
[4] => "nnn ooo ppp"

The basic idea is: a sentence is either inside a pair of brackets, or outside the brackets. A pair of brackets could be nested inside another pair of brackets, but only the most exterior brackets are considered as a sentence.

The following is my code which does not work properly:
<?
$text = "aaa bbb (ccc ddd (eee fff) ggg (xxx yyy zzz)) hhh iii jjj (kkk lll mmm) nnn ooo ppp";
$pattern = "/((.*?))|([^{(^)])+/";}

preg_match_all($pattern, $text, $matches);
var_dump($matches[0]);
?>

Anyone please help me.

Thanks in advance.

bogu

Why dont u use string functions, split when u find open brackets dont split until u find where the breackets close ...

nzhou

bogu wrote:
Why dont u use string functions, split when u find open brackets dont split until u find where the breackets close ...

hi bogu,

Thank you for your response.

But the problem is that brackets could be nested in brackets. Using string split function may not find the most exterior closing bracket. For example,

(ccc ddd (eee fff) ggg (xxx yyy zzz))

I need to split it after "zzz))", not after "fff)"

Could it be possible?

Thanks,

Nathan

NogDog

$text = "aaa bbb (ccc ddd (eee fff) ggg (xxx yyy zzz)) hhh iii jjj (kkk lll mmm) nnn ooo ppp";
$pattern = '/\(((?>[^()]+)|(?R))*\)|[^()]+/';
preg_match_all($pattern, $text, $matches);
print_r($matches[0]);

See the notes on "Recursive patterns" on the PCRE Syntax page.

nzhou

Thank you so much, NogDog.

You save my life. This is a perfect solution. May I ask you what does that mean "?R" in regex?

Thanks again

Nathan

Weedpacket

You'll need something nonstandard (like the recursive patterns NogDog) mentioned if you want to use regular expressions, because regular expressions can't count (and they need to if they want to keep track of how many brackets are open).

The usual method for matching brackets is to start at the left of the string and read each character in turn, keeping count of how many unmatched brackets there are. Each time a '(' shows up, add one to the count; each time a ')' appears, subtract one. If the count is zero and you see a '(', it's the start of an outermost pair of brackets. If the count is one and you see a ')', then that's the end of an outermost pair.

NogDog

nzhou wrote:
Thank you so much, NogDog.

You save my life. This is a perfect solution. May I ask you what does that mean "?R" in regex?

Thanks again

Nathan

It's the recursive conditional pattern indicator - part of the PCRE_EXTENDED options that aren't part of "normal" regexes.