Question

I could not find an optimal solution for this problem. The idea is to change all the urls of a text containing a specific domain with a preg_replace_callback() to base 64 codification. The urls are of this type: http://www.domain.com/?fsdf76sf8sf6fds and separately the other way: http://www.otherdomain.com/file/CA60D10F8ACF7CAA

any ideas for the regex?

Answer 1

What you are looking for is something along the lines of

$s = preg_replace_callback('#(([a-z]+://)|([a-z]+://)?[a-z0-9.-]+\.|\b)domain.com[^\s]+#i', function($match) {
    return base64_encode($match[0]);
}, $string);

That regular expression may be a bit confusing, so let’s break it down:

(  -- domain.com must be preceeded by either
  ([a-z]+://)   -- a protocol such as http://
  |
  ([a-z]+://)?[a-z0-9.-]+\.  -- possibly a protocol and definitely a subdomain
  |
  \b  -- word-break (prevents otherdomain.com from matching!)
)
domain.com  -- the actual domain you're looking for
[^\s]+  -- everything up to the next space (to include path, query string, fragment)

A very simple system to test stuff like this:

<?php

$strings = array(
    // positives
    'a http://www.domain.com/?fsdf76sf8sf6fds z' => 'a xxx z',
    'a www.domain.com/?fsdf76sf8sf6fds z' => 'a xxx z',
    'a http://domain.com/?fsdf76sf8sf6fds z' => 'a xxx z',
    'a domain.com/?fsdf76sf8sf6fds z' => 'a xxx z',
    // negatives
    'a http://www.otherdomain.com/file/CA60D10F8ACF7CAA z' => null,
    'a www.otherdomain.com/file/CA60D10F8ACF7CAA z' => null,
    'a http://otherdomain.com/file/CA60D10F8ACF7CAA z' => null,
    'a otherdomain.com/file/CA60D10F8ACF7CAA z' => null,
);

foreach ($strings as $string => $result) {
    $s = preg_replace_callback('#(([a-z]+://)|([a-z]+://)?[a-z0-9.-]+\.|\b)domain.com[^\s]+#i', function($match) {
        return 'xxx';
    }, $string);

    if (!$result) {
        $result = $string;
    }

    if ($s != $result) {
        echo "FAILED: '$string' got '$s'\n";
    } else {
        echo "OK: '$string'\n";
    }
}

(should you already be unit testing, use that instead, obviously…)