php - why does this regex truncate my string to zero length?

Question

Yesterday I tracked down a strange bug which caused a website display only a white page - no content on it, no error message visible.

I found that a regular expression used in preg_replace was the problem.

I used the regex in order to replace the title html tag in the accumulated content just before echo´ing the html. The html got rather large on the page where the bug occured (60 kb - not too large) and it seemed like preg_replace / the regex used can only handle a string of certain length - or my regex is really messed up (also possible).

Look at this sample program which reproduces the problem (tested on PHP 5.2.9):

---

function replaceTitleTagInHtmlSource($content, $replaceWith) {
  return preg_replace('#(<title>)([\s\S]+)(<\/title>)#i', '$1'.$replaceWith.'$3', $content);
}


$dummyStr = str_repeat('A', 6000);

$totalStr = '<title>foo</title>';

for($i = 0; $i < 10; $i++) {
  $totalStr .= $dummyStr;
}

print 'orignal: ' . strlen($totalStr);
print '<hr />';

$replaced = replaceTitleTagInHtmlSource($totalStr, 'bar');

print 'replaced: ' . strlen($replaced);
print '<hr />';

Output:

orignal: 60018
replaced: 0

So - the function gets a string of length 60000 and returns a string with 0 length. Not what I wanted to do with my regex.

---

Changing

for($i = 0; $i < 10; $i++) {

to

for($i = 0; $i < 1; $i++) {

in order to decrease the total string length, the output is:

orignal: 6018
replaced: 6018

---

When I removed the replacing, the content of the page was displayed without any problems.

Answer 1

Your regex seems to be a little funny.

([\s\S]+) matches all space and non-space. you should try (.*?) instead.

changing your function works for me:

function replaceTitleTagInHtmlSource($content, $replaceWith) {
  return preg_replace('`\<title\>(.*?)\<\/title\>`i', '<title>'.$replaceWith.'</title>', $content);
}

and the problem seems to be you trying to use $1 and $3 to match and

Answer 2

It seems like you're running into the backtracking limit.

This is confirmed if you print preg_last_error(): it returns PREG_BACKTRACK_LIMIT_ERROR.

You can either increase the limit in your ini file or using ini_set() or change your regular expression from ([\s\S]+) to .*?, which will stop it from backtracking so much.

Answer 3

It thas been said many times before on SO, eg http://stackoverflow.com/questions/1458280/regex-to-match-the-first-ending-html-tag (and probably will be mentioned again) that regexes are not appropriate for HTML because tags are too irregular.

Use DOM functions where they're available.

Answer 4

Backtracking: [\s\S]+ will match ALL available characters, then go backwards through the string looking for the </title>. [^<]+ matches all characters that aren't < and therefore grabs </title> faster.

function replaceTitleTagInHtmlSource($content, $replaceWith) {
  return preg_replace('#(<title>)([^<]+)(</title>)#i', '$1'.$replaceWith.'$3', $content);
}