Extract table rows with a specific element from HTML using HTML::TableExtract in perl

I've learned the hard way that regexes cannot adequately parse html, prior to finding post after post about it.

I am trying to extract unread PMs from a webpage that sit in a table. It's the only table on the page being requested, s开发者_开发百科o that part is nice. Each row is a set of columns regarding the PM. The class of the TR informs of an unread/read PM. - which is what is catching me.

I tried to use HTML::TableExtract which almost worked perfectly, except I can't figure out how to check the TR element.

Example Table Structure:

<table>
    <tr class="header">
        <td></td>
        <td>Subject</td>
        <td>Sender</td>
        <td>Date</td>
    </tr>
    <tr class="unread">
        <td>checkbox for multi-edit stuff</td>
        <td>Example of an unread PM</td>
        <td>Me</td>
        <td>Jul 30, 2011</td>
    </tr>
    <tr class="read">
        ....   
    </tr>
</table>

Using HTML::TableExtract I was able to get everything except the unread/read classes. Like so:

$t = HTML::TableExtract->new(keep_html);
$t->parse($lwp_data);
foreach $t2 ($t->tables) {
    foreach $row ($t2->rows) {
#Can't find a way to search for <tr class="unread". As
#Attribute data is stripped at this point by HTML::TableExtract

        #This now shows EVERY PM in the list
        print join(',', @$row), "\n";
    }
 }

How else could I parse this out, and get only the TR's with class="unread"?

Searches resulted in way too complex answers or answers that don't quite solve my problem.

Here's the most recent method I'm using to get what I want (And is working, I just wonder how to do it a better way):

 while ($page =~ m/(unreadpm.*?\/tr)/sg) {
      $data = $1;
      if ($data =~ m(value="(\d+)".*?<a href="(inbox.php\?action=viewconv&amp;id=\d+)">(.*?)</a>\n</strong>\s+</td>\n\s+<td>(.*?)</td>)sg) {
           my ($id,$link,$subject,$user) = ($1, $2, $3, $4);
           if ($user =~ m(user\.php\?id=\d+">(.*?)</a>)) {
                $user = $1;
           }

           if (grep $_ eq $id, @ids) {
                print "Message ID: $id already listed\n"
           } else {
                print "Emailing - Subject: $subject by $user. ID: $id Link: $link ...";
                send_email($subject,$user,$link);
                print "done.\n";
                push @ids, $id;
           }
      }
 }

I can recommend HTML::TreeBuilder in combination with XML::LibXML to do the job.

my $tree = HTML::TreeBuilder->new_from_content( $html );
my $xml  = $tree->as_XML;
my $doc = XML::LibXML->load_xml(string => $xml);

You can then use findvalue to find the <tr> nodes using XPath expressions.

Using HTML::Selector::XPath you can even use CSS selectors to get to the <tr>.

If I've understood the question then I would do something like:

@html_lines = (use curl or otherwise to retrieve the html)

$GET_LINE = 0;

foreach $line (@html_lines)
{
  if ($line =~ /\<tr class="unread"\>/)
  {
      $GET_LINE = 1;
      next;
  }

  if ( ($line =~ |\</tr\>|) && ($GET_LINE) ) 
  { 
      $GET_LINE = 0;
      next;
  }

  if ($GET_LINE)
  {
     #process the <td> lines
  }
}

NB: Im not guaranteeing that syntax is correct, but you get the picture...