From ba46c304bdffdce0b6019a47f6759ce6abbeef90 Mon Sep 17 00:00:00 2001
From: Stefan Ritter <xeno@thehappy.de>
Date: Tue, 16 Aug 2011 10:51:41 +0200
Subject: Initial commit

---
 parse-fileext-wiki.pl | 42 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 42 insertions(+)
 create mode 100644 parse-fileext-wiki.pl

(limited to 'parse-fileext-wiki.pl')

diff --git a/parse-fileext-wiki.pl b/parse-fileext-wiki.pl
new file mode 100644
index 0000000..906751b
--- /dev/null
+++ b/parse-fileext-wiki.pl
@@ -0,0 +1,42 @@
+#!/usr/bin/perl
+
+#use strict;
+#use warnings;
+use Web::Scraper;
+use URI;
+use HTML::Entities;
+use Encode;
+use URI::Escape;
+use LWP::UserAgent;
+
+my $scrap;
+
+my $wikiurl = "http://de.wikipedia.org/wiki/Liste_der_Dateiendungen";
+
+my $scrapp = scraper {
+   process '//div[@id="bodyContent"]/table/tr/td/a', 'chars[]' => 'TEXT';
+};
+my $url = URI->new($wikiurl);
+my $blubb = $scrapp->scrape($url);
+my $list = $blubb->{'chars'};
+
+binmode(STDOUT, ":utf8");
+
+foreach(@$list) {
+   $scrap = scraper {
+      process '//div[@id="bodyContent"]/table[@class="prettytable"]/tr/td', 'table[]' => 'TEXT';
+   };
+   $url = URI->new("$wikiurl/$_");
+
+   my $res = $scrap->scrape($url);
+   my $table = $res->{'table'};
+
+   for ($i=0; $i<=$#$table; $i+=3) {
+      if ($$table[$i] !~ /\..*(\..*)+/ && $$table[$i+1] !~ /^.?$/ ) {
+         print "$$table[$i] $$table[$i+1]\n";
+      }
+      if ($$table[$i+2] =~ /^\./) {
+         $i--;
+      }
+   }
+}
-- 
cgit v1.2.3