Parent Directory
| 
Revision Log
Revision 1.1 - (view) (download) (as text)
| 1 : | olson | 1.1 | |
| 2 : | # package main; | ||
| 3 : | |||
| 4 : | # use Data::Dumper; | ||
| 5 : | # use strict; | ||
| 6 : | |||
| 7 : | # my $p = HtmlSplitter->new(); | ||
| 8 : | # my $ret = $p->parse_file(shift || die); | ||
| 9 : | # print "Done parsing: $ret\n"; | ||
| 10 : | |||
| 11 : | # $p->{head} =~ s/\r\n/\n/gm; | ||
| 12 : | # $p->{head} =~ s/\r/\n/gm; | ||
| 13 : | |||
| 14 : | # $p->{body} =~ s/\r\n/\n/gm; | ||
| 15 : | # $p->{body} =~ s/\r/\n/gm; | ||
| 16 : | |||
| 17 : | # #print "HEAD: $p->{head}\n"; | ||
| 18 : | # #print "BODY: $p->{body}\n"; | ||
| 19 : | |||
| 20 : | # my @maps = @{$p->{map_names}}; | ||
| 21 : | |||
| 22 : | # print "maps: @maps\n"; | ||
| 23 : | # for my $map (@maps) | ||
| 24 : | # { | ||
| 25 : | # print "$map:\n"; | ||
| 26 : | # print $p->{map}->{$map}, "\n"; | ||
| 27 : | |||
| 28 : | # } | ||
| 29 : | |||
| 30 : | package HtmlPageParser; | ||
| 31 : | |||
| 32 : | use strict; | ||
| 33 : | use Data::Dumper; | ||
| 34 : | use HTML::Parser (); | ||
| 35 : | |||
| 36 : | use base 'HTML::Parser'; | ||
| 37 : | |||
| 38 : | sub new | ||
| 39 : | { | ||
| 40 : | my($class) = @_; | ||
| 41 : | |||
| 42 : | my $self = $class->SUPER::new(api_version => 3, | ||
| 43 : | start_h => ["start_handler", "self,tagname,text,attr"], | ||
| 44 : | end_h => ["end_handler", "self,tagname,text,attr"], | ||
| 45 : | text_h => ["text_handler", "self,tagname,text"], | ||
| 46 : | default_h => ["default_handler", "self,text"]); | ||
| 47 : | |||
| 48 : | $self->{state} = 'start'; | ||
| 49 : | return bless($self, $class); | ||
| 50 : | } | ||
| 51 : | |||
| 52 : | sub start_handler | ||
| 53 : | { | ||
| 54 : | my($self, $tag, $txt, $attr) = @_; | ||
| 55 : | |||
| 56 : | # print "Start tag=$tag txt=$txt state=" . $self->state . "\n"; | ||
| 57 : | |||
| 58 : | if ($tag eq 'map') | ||
| 59 : | { | ||
| 60 : | my $name = $attr->{name}; | ||
| 61 : | $self->{in_map} = $name; | ||
| 62 : | push(@{$self->{map_names}}, $name); | ||
| 63 : | } | ||
| 64 : | elsif ($tag eq 'img') | ||
| 65 : | { | ||
| 66 : | my $src = $attr->{src}; | ||
| 67 : | } | ||
| 68 : | |||
| 69 : | if (my $map = $self->{in_map}) | ||
| 70 : | { | ||
| 71 : | $self->{map}->{$map} .= $txt; | ||
| 72 : | } | ||
| 73 : | |||
| 74 : | # | ||
| 75 : | # If we're gathering information from the <HEAD> block, just accumulate text. | ||
| 76 : | # | ||
| 77 : | if ($self->state eq 'head') | ||
| 78 : | { | ||
| 79 : | $self->{head} .= $txt; | ||
| 80 : | } | ||
| 81 : | elsif ($self->state eq 'body') | ||
| 82 : | { | ||
| 83 : | $self->{body} .= $txt; | ||
| 84 : | } | ||
| 85 : | # | ||
| 86 : | # Otherwise, if we see a <head>, start gathering | ||
| 87 : | # | ||
| 88 : | elsif ($tag eq 'head') | ||
| 89 : | { | ||
| 90 : | $self->state('head'); | ||
| 91 : | } | ||
| 92 : | elsif ($tag eq 'body') | ||
| 93 : | { | ||
| 94 : | $self->state('body'); | ||
| 95 : | } | ||
| 96 : | } | ||
| 97 : | |||
| 98 : | sub end_handler | ||
| 99 : | { | ||
| 100 : | my($self, $tag, $txt, $attr) = @_; | ||
| 101 : | |||
| 102 : | if (my $map = $self->{in_map}) | ||
| 103 : | { | ||
| 104 : | $self->{map}->{$map} .= $txt; | ||
| 105 : | } | ||
| 106 : | |||
| 107 : | if ($tag eq 'map') | ||
| 108 : | { | ||
| 109 : | delete $self->{in_map}; | ||
| 110 : | } | ||
| 111 : | |||
| 112 : | |||
| 113 : | # | ||
| 114 : | # If we've finished the head, switch out of head state. | ||
| 115 : | # | ||
| 116 : | if ($tag eq 'head') | ||
| 117 : | { | ||
| 118 : | $self->state('none'); | ||
| 119 : | } | ||
| 120 : | elsif ($tag eq 'body') | ||
| 121 : | { | ||
| 122 : | $self->state('none'); | ||
| 123 : | } | ||
| 124 : | elsif ($self->state eq 'head') | ||
| 125 : | { | ||
| 126 : | $self->{head} .= $txt; | ||
| 127 : | } | ||
| 128 : | elsif ($self->state eq 'body') | ||
| 129 : | { | ||
| 130 : | $self->{body} .= $txt; | ||
| 131 : | } | ||
| 132 : | |||
| 133 : | } | ||
| 134 : | |||
| 135 : | sub text_handler | ||
| 136 : | { | ||
| 137 : | my($self, $tag, $txt) = @_; | ||
| 138 : | |||
| 139 : | # print "txt tag=$tag txt='$txt'\n"; | ||
| 140 : | |||
| 141 : | if (my $map = $self->{in_map}) | ||
| 142 : | { | ||
| 143 : | $self->{map}->{$map} .= $txt; | ||
| 144 : | } | ||
| 145 : | |||
| 146 : | if ($self->state eq 'head') | ||
| 147 : | { | ||
| 148 : | $self->{head} .= $txt; | ||
| 149 : | } | ||
| 150 : | elsif ($self->state eq 'body') | ||
| 151 : | { | ||
| 152 : | $self->{body} .= $txt; | ||
| 153 : | } | ||
| 154 : | |||
| 155 : | } | ||
| 156 : | sub default_handler | ||
| 157 : | { | ||
| 158 : | my($self, $tag, $txt) = @_; | ||
| 159 : | |||
| 160 : | # print "def tag=$tag txt='$txt'\n"; | ||
| 161 : | |||
| 162 : | if (my $map = $self->{in_map}) | ||
| 163 : | { | ||
| 164 : | $self->{map}->{$map} .= $txt; | ||
| 165 : | } | ||
| 166 : | |||
| 167 : | if ($self->state eq 'head') | ||
| 168 : | { | ||
| 169 : | $self->{head} .= $txt; | ||
| 170 : | } | ||
| 171 : | elsif ($self->state eq 'body') | ||
| 172 : | { | ||
| 173 : | $self->{body} .= $txt; | ||
| 174 : | } | ||
| 175 : | } | ||
| 176 : | |||
| 177 : | sub state | ||
| 178 : | { | ||
| 179 : | my($self, $s) = @_; | ||
| 180 : | |||
| 181 : | if (defined($s)) | ||
| 182 : | { | ||
| 183 : | # cluck "set state to $s"; | ||
| 184 : | my $old = $self->{state}; | ||
| 185 : | $self->{state} = $s; | ||
| 186 : | return $old; | ||
| 187 : | } | ||
| 188 : | else | ||
| 189 : | { | ||
| 190 : | return $self->{state}; | ||
| 191 : | } | ||
| 192 : | } | ||
| 193 : | |||
| 194 : | 1; |
| MCS Webmaster | ViewVC Help |
| Powered by ViewVC 1.0.3 |