path: root/posts/2009-06-06-modules-i-like-web-scraper.md

For [$work](http://rtgi.fr) I need to write scrapers. It used to be boring and painful. But thanks to [miyagawa](http://search.cpan.org/~miyagawa/), this is not true anymore. [Web::Scraper](http://search.cpan.org/perldoc?Web::Scraper) offer a nice API: you can write your rules using XPath, you can chaine rules, a nice and simple syntax, etc.

I wanted to export my data from my last.fm account but there is no API for this, so I would need to scrap them. All the data are available [as a web page](http://www.last.fm/user/franckcuny/tracks) that list your music. So the scraper need to find how many pages, and find the content on each page to extract a list of your listening.

For the total of pages, it's easy. Let's take a look at the HTML code and search for something like this:

``` html
    <a class="lastpage" href="/user/franckcuny/tracks?page=272">272</a>
```

the information is in a class **lastpage**.

Now we need to find our data: I need the artist name, the song name and the date I played this song.

All this data are in a **table**, and each new entry is in a **td**.

``` html
    <tr id="r9_1580_1920248170" class="odd">
    [...]
        <td class="subjectCell">
        <a href="/music/Earth">Earth</a>
        <a href="/music/Earth/_/Sonar+and+Depth+Charge">Sonar and Depth Charge</a>
        </td>
    [...]
    <td class="dateCell last">
        <abbr title="2009-05-13T15:18:25Z">13 May 3:18pm</abbr>
    </td>
```

It's simple: information about a song are stored in **subjectcell**, and the artist and song title are each in a tag **a**. The date is in a **dateCell**, and we need the **title** from the **abbr** tag.

The scraper we need to write is

``` perl
my $scrap = scraper {
    process 'a[class="lastpage"]', 'last'    => 'TEXT';
    process 'tr',                  'songs[]' => scraper {
        process 'abbr',                    'date' => '@title';
        process 'td[class="subjectCell"]', 'song' => scraper {
            process 'a', 'info[]' => 'TEXT';
        };
    }
};
```

The first rule extract the total of page. The second iter on each **tr** and store the content in an array named **songs**. This **tr** need to be scraped. So we look the the **abbr** tag, and store in **date** the property **title**. Then we look for the song and artitst information. We look for the **td** with a class named **subjectCell**, a extract all links.

Our final script will look like this:

``` perl
 #!/usr/bin/perl -w
 use strict;
 use feature ':5.10';

 use Web::Scraper;
 use URI;
 use IO::All -utf8;

 my $username = shift;
 my $output   = shift;

 my $scrap = scraper {
     process 'a[class="lastpage"]', 'last'    => 'TEXT';
     process 'tr',                  'songs[]' => scraper {
         process 'abbr',                    'date' => '@title';
         process 'td[class="subjectCell"]', 'song' => scraper {
             process 'a', 'info[]' => 'TEXT';
         };
     }
 };

 my $url = "http://www.last.fm/user/" . $username . "/tracks?page=";
 scrap_lastfm(1);

 sub scrap_lastfm {
     my $page      = shift;
     my $scrap_uri = $url . $page;
     say $scrap_uri;
     my $res      = $scrap->scrape(URI->new($scrap_uri));
     my $lastpage = $res->{last};
     foreach my $record (@{$res->{songs}}) {
         my $line = join("\t", @{$record->{song}->{info}}, $record->{date});
         $line . "\n" >> io $output;
     }
     $page++;
     scrap_lastfm($page) if $page <= $lastpage;
}
```

You can use this script like this:

``` bash
perl lastfmscraper.pl franckcuny store_data.txt
```