Question

the following is a simplification of a regex i am using. on my development machine both $pattern1 and $pattern2 return a match, however on my production machine only $pattern1 returns a match! clearly the only difference between $pattern1 and $pattern2 is that one of them has brackets around a word. however both are valid patterns which should match the given haystack (as far as i know).

$pattern1 = '/\<a name="ERROR TEXT"\>\<\/a\>\s*?validated\s*?\<\/span\>\s*?\<\/h1\>/';
$pattern2 = '/\<a name="ERROR TEXT"\>\<\/a\>\s*?(validated)\s*?\<\/span\>\s*?\<\/h1\>/';
$haystack = '- IFCS msg value, BOOKMARKED AS ERROR TEXT -->
          <a name="ERROR TEXT"></a>
             validated</span>
       </h1>

                <!-- START: .formActionHolder -->
                <div class="formActionHolder">';
preg_match($pattern1, $haystack, $matches);
print_r($matches);

has anyone found this problem before? note that this is not the whole of the regex – this is a simplified version which i have identified as being the problem. in my actual code, the value of ‘validated’ is not a constant – hence my reason for using brackets to capture the word. of course the patterns have other characters within the parenthesis as well so that i can capture the variable words here. this is just a simplified example which hones in on the problem that i am having with two seemingly fine regexes.

on my development machine i am using php5.3.2 with the pcre 7.8 library and on my production machine i am using php5.2.4 with pcre 7.4.

Answer 1

i had a thought about the ?( combination in $pattern2 so i removed the ? to make

$pattern = '/\<a name="ERROR TEXT"\>\<\/a\>\s*(validated)\s*?\<\/span\>\s*?\<\/h1\>/';

and that works!! its very strange – possibly even a bug?

so it looks like the ?(validated) bit was being interpreted as a conditional subpattern rather than the question mark being used to make the \s* ungreedy

that doesn’t look like correct behavior to me.

ah well…its a bit of a pain since now my * will be greedy. the regex pattern does what i want in this instance though…

thanks for all your helpful comments!