1 |
#!/usr/local/bin/perl -w |
2 |
use strict; |
3 |
|
4 |
use LWP::UserAgent; |
5 |
use HTTP::Status; |
6 |
use HTML::LinkExtor; |
7 |
|
8 |
if (scalar(@ARGV) != 2) { |
9 |
print STDERR "Usage: SwishSpider localpath url\n"; |
10 |
exit(1); |
11 |
} |
12 |
|
13 |
my $ua = new LWP::UserAgent; |
14 |
$ua->agent( "SwishSpider http://swish-e.org" ); |
15 |
|
16 |
my $localpath = shift; |
17 |
my $url = shift; |
18 |
|
19 |
my $no_parent_url; |
20 |
if ($url =~ m/\s/) { |
21 |
($url,$no_parent_url) = split(/\s/,$url,2); |
22 |
} |
23 |
|
24 |
my $request = new HTTP::Request( "GET", $url ); |
25 |
my $response = $ua->simple_request( $request ); |
26 |
|
27 |
# |
28 |
# Write out important meta-data. This includes the HTTP code. Depending on the |
29 |
# code, we write out other data. Redirects have the location printed, everything |
30 |
# else gets the content-type. |
31 |
# |
32 |
open( RESP, ">$localpath.response" ) || die( "Could not open response file $localpath.response" ); |
33 |
|
34 |
print RESP $response->code() . "\n"; |
35 |
if( $response->code() == RC_OK ) { |
36 |
print RESP $response->header( "content-type" ) . "\n"; |
37 |
} elsif( $response->is_redirect() ) { |
38 |
my $link = $response->header( "location" ); |
39 |
if ($no_parent_url) { |
40 |
if ($link =~ m/$no_parent_url/) { |
41 |
# if this URL is below parent URL o.k.... |
42 |
print RESP "$link $no_parent_url\n"; |
43 |
} else { |
44 |
# if not, crawl just this page! |
45 |
print RESP "$link $link\n"; |
46 |
} |
47 |
} else { |
48 |
print RESP "$link\n"; |
49 |
} |
50 |
} |
51 |
close( RESP ); |
52 |
|
53 |
# |
54 |
# Write out the actual data assuming the retrieval was succesful. Also, if |
55 |
# we have actual data and it's of type text/html, write out all the links it |
56 |
# refers to |
57 |
# |
58 |
if( $response->code() == RC_OK ) { |
59 |
my $contents = $response->content(); |
60 |
|
61 |
open( CONTENTS, ">$localpath.contents" ) || die( "Could not open contents file $localpath.contents\n" ); |
62 |
# if you don't want content to be indexed, include it in |
63 |
# <noindex> foobar </noindex> tags or surround it with comments |
64 |
# <!-- noindex --> foobar <!-- /noindex --> |
65 |
$contents =~ s,<noindex>.+?</noindex>,,isg; |
66 |
$contents =~ s,<!--\s*noindex\s*-->.+?<!--\s*/noindex\s*-->,,isg; |
67 |
# this will remove all script from indexing content |
68 |
$contents =~ s,<script>.+?</script>,,isg; |
69 |
# remap Windows charset to ISO-8859-2 |
70 |
$contents =~ tr/ðèæÐÈÆ/¹ð¾èæ©Ð®ÈÆ/; # 1250 -> iso8859-2 |
71 |
# this will fix badly formatted html in form: |
72 |
# <head><title>some text</title |
73 |
# ></head> |
74 |
# which will confuse indexer (or libxml2?) |
75 |
$contents =~ s/[\n\r]^(>)/$1\n/msg; |
76 |
print CONTENTS $contents; |
77 |
close( CONTENTS ); |
78 |
|
79 |
if( $response->header("content-type") =~ "text/html" ) { |
80 |
open( LINKS, ">$localpath.links" ) || die( "Could not open links file $localpath.links\n" ); |
81 |
my $p = HTML::LinkExtor->new( \&linkcb, $url ); |
82 |
$p->parse( $contents ); |
83 |
|
84 |
close( LINKS ); |
85 |
} |
86 |
} |
87 |
|
88 |
|
89 |
sub linkcb { |
90 |
my($tag, %links) = @_; |
91 |
if (($tag eq "a" || $tag eq "area") && ($links{"href"})) { |
92 |
my $link = $links{"href"}; |
93 |
|
94 |
# |
95 |
# Remove fragments |
96 |
# |
97 |
$link =~ s/(.*)#.*/$1/; |
98 |
|
99 |
# |
100 |
# Remove ../ This is important because the abs() function |
101 |
# can leave these in and cause never ending loops. |
102 |
# |
103 |
$link =~ s/\.\.\///g; |
104 |
|
105 |
# hack for apostrophe -- changes URL, but should work for most clients. |
106 |
$link =~ s/'/%27/g; |
107 |
|
108 |
# hack for Apache directory listings |
109 |
$link =~ s,/\?[NMSD]=[AD]$,/,g; |
110 |
|
111 |
# speedup, skip pictures |
112 |
return if ($link =~ m/\.(gif|jpg|png)/); |
113 |
|
114 |
if ($no_parent_url) { |
115 |
if ($link =~ m/$no_parent_url/) { |
116 |
print LINKS "$link $no_parent_url\n"; |
117 |
# print STDERR "using $link\n"; |
118 |
# } else { |
119 |
# print STDERR "skipping $link\n"; |
120 |
} |
121 |
} else { |
122 |
print LINKS "$link\n"; |
123 |
} |
124 |
} |
125 |
} |
126 |
|