正则表达式在多行上太贪心了

Question

I have the following code:

$text = "Lorem ipsum dolor sit amet, [b]consectetur adipiscing elit[/b]. 
Nunc lorem velit, lacinia ut commodo in, suscipit vitae magna. 
Nam imperdiet neque blandit semper tempus. 
Curabitur sapien ante, vestibulum vitae ante a, condimentum dignissim tortor. Aenean adipiscing tincidunt lorem, non eleifend tellus suscipit a. Nulla convallis [b]
pulvinar ligula[/b], at tempor ante. Fusce a tellus enim. Vivamus nibh eros, ultrices at auctor quis, fringilla nec dolor. Aenean nec tincidunt odio, id pulvinar felis. Pellentesque in augue volutpat, gravida nibh eu, lobortis augue.";

preg_match_all("#(\[b\].*\[/b\])#s", $text, $value);

my $value is returning from the first [b] to the last [/b]. I need it to match each pair individually.

As I understand it, I have to use the s at the end to select multiple lines, but the * is then being too greedy. I can't use just a ? as I the number of characters can vary... what am I missing?

Answer 1

This is a common mistake. Unless you do something to avoid it, the regex engine will find the longest substring that can possibly be matched by your pattern. Depending on the context, there might be various possible solutions, but for engines that support Perl regex syntaxes, the easiest is generally to use the "non-greedy" variant of the repetition operator you are using. That is, *? instead of * , +? instead of + , ?? instead of ? or {m,n}? instead of {m,n} .

So in your example, the pattern should read as:

preg_match_all("#(\[b\].*?\[/b\])#s", $text, $value);

Answer 2

Another way to avoid the lazy quantifier:

preg_match_all('~\[b](?>[^[]++|\[(?!/b]))*+\[/b]~', $text, $value);

With this way, you avoid two problems:

1. greedy quantifier is not a problem, since the character class stop at each opening square bracket
2. since you don't use the dot, you don't care about the 's' modifier and newlines.