How would one use PHP preg_match_all to differentiate anchor elements identified by attribute of inner HTML element?

Question

I have sets of HTML anchor elements enclosing image elements. For each set, using PHP-CLI, I want to pull the URLs and classify them according to their types. The type of anchor can only be determined by an attribute of its child image element. It would be easy if there was only one of each type per set. My problem is when two anchor elements of one type are separated by one or more of the other types. My non-greedy parenthesized sub-pattern seems to become greedy and expands to find the second relevant child attribute. In my test script I'm trying to pull the 'Userlink' URLs from amongst the other types. Using a simple pattern like:

#<a href="(.*?)" custattr="value1"><img alt="Userlink"#

On a set like:

<li><a href="http://www.userlink1.com/my/page.html" custattr="value1"><img alt="Userlink" class="common_link_class" height="123" src="pic0.png" width="123" style="width: 123px;"></a></li><li><a href="http://www.socnet1.com/username1" custattr="value1"><img alt="Socnet1" class="common_link_class" height="123" src="pic1.png" width="123" style="width: 123px;"></a></li><li><a href="http://www.socnet2.com/username1" custattr="value1"><img alt="Socnet2" class="common_link_class" height="123" src="pic2.png" width="123" style="width: 123px;"></a></li><li><a href="mailto:useralias1@unlikely.zyx321.usermail.net" custattr="value1"><img alt="Usermail" class="common_link_class" height="123" src="pic3.png" width="123" style="width: 123px;"></a></li><li><a href="http://www.userlink2.com/my/page.html" custattr="value1"><img alt="Userlink" class="common_link_class" height="123" src="pic4.png" width="123" style="width: 123px;"></a></li>

(sorry, but the actual html is on one line like that)

My sub-pattern captures from the beginning of the first "Userlink" URL to the end of the last one.

I've tried many variations of look-aheads, not sure I should list them all here. So far they've either returned no match at all or the same as described above.

Here's my test script (running in a Bash shell):

#!/usr/bin/php
<?
    $lines = 0;
    $input = "";
    $matches = array();

    while ($line = fgets(STDIN)){
        $input .= $line;
        $lines++;
    }
    fwrite(STDERR, "Processing $lines\n");

    $pcre = '#<a href="(.*?)" custattr="value1"><img alt="Userlink"#';

    if (preg_match_all($pcre,$input,$matches)){
        fwrite(STDERR, "\$matches has " . count($matches) . " elements\n");
        foreach ($matches[1] as $match){
            fwrite(STDOUT, $match . "\n");
        }
    }
?>

What PCRE pattern for PHP's preg_match_all() would return the two "Userlink" URLs in the above example?

[**Don't parse HTML with regex**](http://stackoverflow.com/a/1732454/2057919). Use a parser. — elixenide, Feb 27 '14 at 20:11
instead of using an ungreedy `.*?` use the greedy character class `[^"]*`. — Casimir et Hippolyte, Feb 27 '14 at 20:11
As said Ed Cottrell with this *^?!# link, if you only want to find href content, using the DOM can be a good option. — Casimir et Hippolyte, Feb 27 '14 at 20:17
Even if I don't need to identify or use the HTML elements per se, and am throwing away all of them, would an HTML parser still be better? — TrooPhalce, Feb 27 '14 at 23:19

score 0 · Answer 1 · answered Feb 27 '14 at 20:31

This regex should work -

<a href="([^"]*?)"[^>]*\><img alt="Userlink"

You can see how it work here.

Testing it -

$pcre = '/<a href="([^"]*?)"[^>]*\><img alt="Userlink"/';
if (preg_match_all($pcre,$input,$matches)){
    var_dump($matches);
    //$matches[1] will be the array containing the urls.
}
/*
    OUTPUT- 
    array
      0 => 
        array
          0 => string '<a href="http://www.userlink1.com/my/page.html" custattr="value1"><img alt="Userlink"' (length=85)
          1 => string '<a href="http://www.userlink2.com/my/page.html" custattr="value1"><img alt="Userlink"' (length=85)
      1 => 
        array
          0 => string 'http://www.userlink1.com/my/page.html' (length=37)
          1 => string 'http://www.userlink2.com/my/page.html' (length=37)
*/

score 0 · Accepted Answer · answered Feb 27 '14 at 20:31

I have taken the liberty of changing your variable names:

$pattern = '~<a href="([^"]++)" custattr="value1"><img alt="Userlink"~';

if ($nb = preg_match_all($pattern, $input, $matches)) {
    fwrite(STDERR, "\$matches has " . $nb . " elements\n");
    fwrite(STDOUT, implode("\n", $match) . "\n");
}

Note that the preg_match_all function returns the number of matches.

How would one use PHP preg_match_all to differentiate anchor elements identified by attribute of inner HTML element?

2 Answers2