|
@@ -1,19 +1,29 @@
|
1
|
1
|
#!/usr/bin/perl
|
2
|
2
|
use strict;
|
3
|
3
|
use Digest::SHA qw(sha256_hex);
|
4
|
|
-use LWP::UserAgent;
|
|
4
|
+use HTTP::Tiny;
|
5
|
5
|
use DBI qw(:sql_types);
|
6
|
6
|
use Getopt::Long;
|
7
|
|
-use pQuery;
|
|
7
|
+use Mojo::DOM;
|
|
8
|
+use HTML::WikiConverter;
|
|
9
|
+use Switch;
|
|
10
|
+#use Data::Dumper;
|
8
|
11
|
|
9
|
|
-sub init {
|
10
|
|
- my $driver = "SQLite";
|
11
|
|
- my $database = "history.db";
|
|
12
|
+sub trim {
|
|
13
|
+ (my $s = $_[0]) =~ s/^\s+|\s+$//g;
|
|
14
|
+ return $s;
|
|
15
|
+}
|
|
16
|
+
|
|
17
|
+sub connect_db($$$){
|
|
18
|
+ my ($driver, $database, $password) = @_;
|
12
|
19
|
my $dsn = "DBI:$driver:dbname=$database";
|
13
|
20
|
my $dbh = DBI->connect($dsn, { RaiseError => 1 })
|
14
|
21
|
or die $DBI::errstr;
|
15
|
22
|
print "Opened database successfully\n";
|
16
|
|
- return $dbh;
|
|
23
|
+}
|
|
24
|
+
|
|
25
|
+sub init {
|
|
26
|
+ return connect_db("SQLite", "history.db", "");
|
17
|
27
|
}
|
18
|
28
|
|
19
|
29
|
sub add_History($$$){
|
|
@@ -23,14 +33,20 @@ sub add_History($$$){
|
23
|
33
|
my $rv = $dbh->do($stmt) or die $DBI::errstr;
|
24
|
34
|
}
|
25
|
35
|
|
26
|
|
-sub get_html($) {
|
27
|
|
- my $ua = new LWP::UserAgent;
|
28
|
|
- $ua->timeout(120);
|
29
|
|
- my $url = @_ ;
|
30
|
|
- my $request = new HTTP::Request('GET', $url);
|
31
|
|
- my $response = $ua->request($request);
|
32
|
|
- my $content = $response->content();
|
33
|
|
- return $content;
|
|
36
|
+sub get_html($){
|
|
37
|
+ my ($url) = @_;
|
|
38
|
+ my $response = HTTP::Tiny->new->get($url);
|
|
39
|
+ if ($response->{success}) {
|
|
40
|
+ return $response->{content};
|
|
41
|
+ }
|
|
42
|
+ return 0;
|
|
43
|
+}
|
|
44
|
+
|
|
45
|
+sub get_content($$){
|
|
46
|
+ my($url, $selector) = @_;
|
|
47
|
+ my $content = get_html( $url );
|
|
48
|
+ my $dom = Mojo::DOM->new( $content );
|
|
49
|
+ return $dom->at( "$selector" );
|
34
|
50
|
}
|
35
|
51
|
|
36
|
52
|
sub visit($$){
|
|
@@ -41,15 +57,19 @@ sub visit($$){
|
41
|
57
|
|
42
|
58
|
sub isNew($$$){
|
43
|
59
|
my($url, $signature, $dbh) = @_;
|
44
|
|
- my $sth = $dbh->prepare("SELECT COUNT(signature) FROM visit WHERE signature=?1 AND url=?2");
|
45
|
|
- $sth->execute($signature, $url);
|
46
|
|
- my $refs = $sth->fetchall_arrayref()->[0][0];
|
47
|
|
- return ($refs)? 0 : 1 ;
|
|
60
|
+ # my $sth = $dbh->prepare("SELECT count(signature) AS COUNT FROM visit WHERE signature=?1 AND url=?2;");
|
|
61
|
+ my $sth = $dbh->prepare("SELECT count(signature) AS COUNT FROM visit WHERE signature='$signature' AND url='$url';");
|
|
62
|
+ #$sth->execute("$signature", "$url");
|
|
63
|
+ $sth->execute();
|
|
64
|
+ my $refs = $sth->fetchrow_arrayref()->[0];
|
|
65
|
+ print "SELECT count(signature) AS COUNT FROM visit WHERE signature=$signature AND url=$url\n";
|
|
66
|
+ print "$url, count : $refs\n";
|
|
67
|
+ return !($refs);
|
48
|
68
|
}
|
49
|
69
|
|
50
|
|
-sub registerIfNew($$){
|
51
|
|
- my($url, $dbh) = @_;
|
52
|
|
- my $signature = visit($url, $dbh);
|
|
70
|
+sub registerIfNew($$$){
|
|
71
|
+ my($url, $selector,$dbh) = @_;
|
|
72
|
+ my $signature = sha256_hex( get_content($url, $selector) );
|
53
|
73
|
|
54
|
74
|
if( isNew($url, $signature ,$dbh)){
|
55
|
75
|
add_History($url, $signature ,$dbh);
|
|
@@ -58,32 +78,55 @@ sub registerIfNew($$){
|
58
|
78
|
return 0;
|
59
|
79
|
}
|
60
|
80
|
|
61
|
|
-sub importFromUrl($){
|
62
|
|
- my ($url) = @_;
|
63
|
|
- print "importing $url";
|
|
81
|
+sub convertStrategy($$$){
|
|
82
|
+ my ($content, $fromFormat, $toFormat) = @_;
|
|
83
|
+
|
|
84
|
+ switch("$fromFormat $toFormat") {
|
|
85
|
+ case "html md" { return htmlToMd( $content ) }
|
|
86
|
+ case "a" { print "string a" }
|
|
87
|
+ else { return $content }
|
|
88
|
+ }
|
|
89
|
+}
|
|
90
|
+
|
|
91
|
+sub importContent($$$$){
|
|
92
|
+ my ($url, $selector, $fromFormat, $toFormat) = @_;
|
|
93
|
+ my $content = get_content( $url, $selector );
|
|
94
|
+
|
|
95
|
+ if( $content ){
|
|
96
|
+ return convertStrategy( $content, $fromFormat, $toFormat );
|
|
97
|
+ }
|
|
98
|
+ return 0;
|
64
|
99
|
}
|
65
|
100
|
|
66
|
|
-sub checkUrl($){
|
67
|
|
- my($url) = @_;
|
|
101
|
+sub checkUrl($$){
|
|
102
|
+ my($url, $selector) = @_;
|
68
|
103
|
my $dbh = init;
|
69
|
|
- if( registerIfNew($url, $dbh) ){
|
70
|
|
- importFromUrl($url);
|
|
104
|
+ if( registerIfNew($url, $selector,$dbh) ){
|
|
105
|
+ get_content( $url, $selector )
|
71
|
106
|
}
|
72
|
107
|
$dbh->disconnect();
|
73
|
108
|
}
|
74
|
109
|
|
75
|
|
-sub get_content($$){
|
76
|
|
- ;
|
|
110
|
+sub htmlToMd($){
|
|
111
|
+ my($html) = @_;
|
|
112
|
+ my $wc = new HTML::WikiConverter( dialect => 'Markdown' );
|
|
113
|
+ return $wc->html2wiki( html => $html );
|
77
|
114
|
}
|
78
|
115
|
|
79
|
116
|
sub main($){
|
|
117
|
+ my $dbh = init;
|
80
|
118
|
my($file) = @_;
|
81
|
|
- print "Using file $file\n";
|
82
|
|
- #checkUrl( $url );
|
83
|
119
|
open(FH, '<', $file) or die $!;
|
84
|
120
|
while(<FH>){
|
85
|
|
- print "Checking $_";
|
86
|
|
- checkUrl( $_ );
|
|
121
|
+ my($url, $selector, $fromFormat, $toFormat) = split("\t");
|
|
122
|
+ $fromFormat = trim($fromFormat);
|
|
123
|
+ $toFormat = trim($toFormat);
|
|
124
|
+ #print get_content( $url, $selector );
|
|
125
|
+ #checkUrl( $_ );
|
|
126
|
+ #print registerIfNew( $url, $selector, $dbh );
|
|
127
|
+ print " $url, $selector, $fromFormat, $toFormat ";
|
|
128
|
+ print importContent( $url, $selector, $fromFormat, $toFormat );
|
|
129
|
+ print "\n"
|
87
|
130
|
}
|
88
|
131
|
close(FH);
|
89
|
132
|
}
|