home | O'Reilly's CD bookshelfs | FreeBSD | Linux | Cisco | Cisco Exam  


Book HomePerl & LWPSearch this book

9.5. Example: Fresh Air

Another HTML::TokeParser problem (in Chapter 8, "Tokenizing Walkthrough") was extracting relevant links from the program descriptions from the Fresh Air web site. There were aspects of the task that we will not review here (such as how to request a month's worth of weekday listings at a time), but we will instead focus on the heart of the program, which is how to take HTML source from a local file, feed it to HTML::TreeBuilder, and pull the interesting links out of the resulting tree.

If we save the HTML source of a program description page as fresh1.html and sift through its source, we get a 12-KB file. Only about one 1 KB of that is real content, like this:

...
<A HREF="http://www.npr.org/ramfiles/fa/20010702.fa.ram">
  <FONT FACE="Verdana, Charcoal, Sans Serif" COLOR="#FFCC00" SIZE="2">
    Listen to <B>Monday - July 2, 2001</B>
  </FONT>
</A>
 
...
 
   <A HREF="http://www.npr.org/ramfiles/fa/20010702.fa.01.ram">Listen to
   <FONT FACE="Verdana, Charcoal, Sans Serif" COLOR="#ffffff" SIZE="3">
   <B> Editor and writer Walter Kirn                            </B>
   </FONT></A>
                             
<BR>
<FONT FACE="Verdana, Charcoal, Sans Serif" COLOR="#ffffff" SIZE="2">
<BLOCKQUOTE>Editor and writer <A
HREF="http://freshair.npr.org/guestInfoFA.cfm?name=walterkirn">Walter
Kirn</A>'s new novel <I>Up in the Air</I> (Doubleday) is about 
...
</BLOCKQUOTE></FONT>
<BR>
 
  <A HREF="http://www.npr.org/ramfiles/fa/20010702.fa.02.ram">Listen to
  <FONT FACE="Verdana, Charcoal, Sans Serif" COLOR="#ffffff" SIZE="3">
  <B> Casting director and actress Joanna Merlin             </B>
  </FONT></A>
 
<BR>
<FONT FACE="Verdana, Charcoal, Sans Serif" COLOR="#ffffff" SIZE="2">
<BLOCKQUOTE>Casting director and actress <A
HREF="http://freshair.npr.org/guestInfoFA.cfm?name=joannamerlin">Joanna
Merlin</A> has written a new guide for actors, <I>Auditioning: An
...
</BLOCKQUOTE></FONT>
<BR>
...

The rest of the file is mostly taken up by some JavaScript, some search box forms, and code for a button bar, which contains image links like this:

...
<A HREF="dayFA.cfm?todayDate=archive"><IMG SRC="images/nav_archived_on.gif" 
ALT="Archived Shows" WIDTH="124" HEIGHT="36" BORDER="0" HSPACE="0" VSPACE="0"></A>
<A HREF="commFA.cfm"><IMG SRC="images/nav_commentators_off.gif" ALT="Commentators" 
WIDTH="124" HEIGHT="36" BORDER="0" HSPACE="0" VSPACE="0"></A>
<A HREF="aboutFA.cfm"><IMG SRC="images/nav_about_off.gif" ALT="About Fresh Air" 
WIDTH="124" HEIGHT="36" BORDER="0" HSPACE="0" VSPACE="0"></A>
<A HREF="stationsFA.cfm"><IMG SRC="images/nav_stations_off.gif" ALT="Find a Station" 
WIDTH="124" HEIGHT="36" BORDER="0" HSPACE="0" VSPACE="0"></A>
...

Then, after the real program description text, there is code that links to the description pages for the previous and next shows:

...
<TD WIDTH="50%" ALIGN="left" BGCOLOR="#4F4F85">
  <FONT FACE="Verdana, Charcoal, Sans Serif" SIZE="2" COLOR="#FFCC00">
    &#160;&#160;&#171;&#160;
  </FONT>
  <A HREF="dayFA.cfm?todayDate=06%2F29%2F2001">
    <FONT FACE="Verdana, Charcoal, Sans Serif" SIZE="2" COLOR="#FFCC00">
      Previous show
    </FONT>
 </A>
</TD>
<TD WIDTH="50%" ALIGN="right" BGCOLOR="#4F4F85">
  <A HREF="dayFA.cfm?todayDate=07%2F03%2F2001">
    <FONT FACE="Verdana, Charcoal, Sans Serif" SIZE="2" COLOR="#FFCC00">
      Next show
    </FONT>
  </A>
  <FONT FACE="Verdana, Charcoal, Sans Serif" SIZE="2" COLOR="#FFCC00">
   &#160;&#187;&#160;&#160;
  </FONT>
</TD>
...

The trick is in capturing the URLs and link text from each program link in the main text, while ignoring the button bar links and the "Previous Show" and "Next Show" links. Two criteria distinguish the links we want from the links we don't: First, each link that we want (i.e., each a element with an href attribute) has a font element as a child; and secondly, the text content of the a element starts with "Listen to" (which we incidentally want to leave out when we print the link text). This is directly implementable with calls to HTML::Element methods:

use HTML::TreeBuilder;
my $tree = HTML::TreeBuilder->new;
$tree->parse_file( 'fresh1.html' ) || die $!;
my $base_url = 'http://www.freshair.com/whatever';
  # for resolving relative URLs

foreach my $a ( $tree->find_by_tag_name('a') ) {

  my $href = $a->attr('href') || next;
    # Make sure it has an href attribute

  next unless grep ref($_) && $_->tag eq 'font', $a->content_list;
    # Make sure (at least) one of its children is a font element
  
  my $text_content = $a->as_text;
  next unless $text_content =~ s/^\s*Listen to\s+//s;
    # Make sure its text content starts with that (and remove it)

  # It's good!  Print it:
  use URI;
  print "$text_content\n  ", URI->new_abs($href, $base_url), "\n";
}

$tree->delete;  # Delete tree from 
memory


Library Navigation Links

Copyright © 2002 O'Reilly & Associates. All rights reserved.