Three bugs.

preg_replace:

$text = preg_replace('/(foo(bar)?) is a good word/',
                     'wibble', $text);

Simple enough.  How about:

$text = preg_replace(/'(foo(bar)?) is a good word/e',
                     '(length(\'\2\')>0)?"wibble":"wobble"',
                     $text);

The first thing to note here is that the idea of substituting into the replacement string like this was a very bad idea, I would encourage you to phase this out in favour of $<num> replacement.

The two obvious things that PHP could get wrong with this form of substitution, PHP gets wrong :-)

Firstly - when \2 does not exist because there was no match, you should should get '', infact with PHP you get ^B, it seems you're simply looking for \<nums> that created matches rather than all \<nums.

Secondly - as a security-aware person, I immediate recognise the problems that '\1' could cause.  A quick look at the code reveals that (thankfully) some effort is being made to quote the inserted string (undocumentedly).  However, the code in PHP uses addslashes() which was designed for database use and not internal PHP single-quote escaping.  PHP's single-quotes only look for \' and \\ and therefore the escaping of " to \" and NULL to \0 in addslashes() will cause spurious backslashes to enter the text.

On an aside note, I also think it was a bad idea to put delimiters into the search string, there is no point to this at all and is just a burden to the user.

PHP does not support all of perl's delimiters, particularly it does not support the (), {}, [] matching delimiters.  This code will not work:

preg_replace("{wibble}", "wobble", $text);