From: mitty <mitty@7d2118f6-f56c-43e7-95a2-4bb3031d96e7>
Date: Mon, 22 Apr 2013 06:50:49 +0000 (+0000)
Subject:  * crawler for Nikkei editorial articles
X-Git-Url: http://lab.mitty.jp/git/?a=commitdiff_plain;ds=inline;h=151083b801351c91b78545e521da845c1b139530;p=lab.git

 * crawler for Nikkei editorial articles

git-svn-id: https://lab.mitty.jp/svn/lab/trunk@213 7d2118f6-f56c-43e7-95a2-4bb3031d96e7
---

diff --git a/misc/nikkei.pl b/misc/nikkei.pl
new file mode 100644
index 0000000..1df153d
--- /dev/null
+++ b/misc/nikkei.pl
@@ -0,0 +1,37 @@
+#! /usr/bin/env perl -w
+
+use strict;
+use warnings;
+use utf8;
+
+use LWP::Simple;
+binmode STDOUT => 'encoding(utf8)';
+
+my $directory = shift @ARGV || "./";
+my $nikkei_url = 'http://www.nikkei.com/news/editorial/';
+my $nikkei     = 'http://www.nikkei.com';
+
+my $regex = 'href="([^"]+)(DGXDZO\w+000)/';
+my $javascript = '<script .*?</script>';
+
+my $content = get($nikkei_url);
+while ($content =~ /$regex/g) {
+    my $article = "$nikkei$1$2/";
+    my $file = "$directory/$2.html";
+    
+    if (-f "$file") { next; }
+    
+    system("wget", "-q", $article, '-O', "$file");
+    sleep 1;
+    
+    if (-f "$file") {
+        open my $html, "<", $file;
+        local $/;
+        my $body = <$html>;
+        
+        $body =~ s/$javascript//g;
+        
+        open $html, ">", $file;
+        print $html $body;
+    }
+}