[7492] | 1 |
|
---|
| 2 | sub readin_html
|
---|
| 3 | {
|
---|
| 4 | my ($html_fname) = @_;
|
---|
| 5 |
|
---|
| 6 | open(HIN,"<$html_fname")
|
---|
| 7 | || die "Unable to open $html_fname: $!\n";
|
---|
| 8 |
|
---|
| 9 | my $html_text;
|
---|
| 10 | my $line;
|
---|
| 11 | while (defined ($line=<HIN>)) {
|
---|
| 12 | $html_text .= $line;
|
---|
| 13 | }
|
---|
| 14 | close(HIN);
|
---|
| 15 |
|
---|
| 16 | return $html_text;
|
---|
| 17 | }
|
---|
| 18 |
|
---|
| 19 | sub stripout_anchortags
|
---|
| 20 | {
|
---|
| 21 | my ($html_text) = @_;
|
---|
| 22 |
|
---|
| 23 | my @anchor_tags = ($html_text =~ m/(<a\s+.*?>)+/gs);
|
---|
| 24 |
|
---|
| 25 | return @anchor_tags;
|
---|
| 26 | }
|
---|
| 27 |
|
---|
| 28 |
|
---|
| 29 | sub print_tags
|
---|
| 30 | {
|
---|
| 31 | my (@tags) = @_;
|
---|
| 32 |
|
---|
| 33 | my $a;
|
---|
| 34 | foreach $a ( @tags) {
|
---|
| 35 | print "$a\n";
|
---|
| 36 | }
|
---|
| 37 | }
|
---|
| 38 |
|
---|
| 39 | sub filter_tags
|
---|
| 40 | {
|
---|
| 41 | my ($filter_text,@tags) = @_;
|
---|
| 42 |
|
---|
| 43 | my @filtered_tags = ();
|
---|
| 44 |
|
---|
| 45 | my $t;
|
---|
| 46 | foreach $t (@tags) {
|
---|
| 47 | if ($t =~ m/$filter_text/x) {
|
---|
| 48 | push(@filtered_tags,$t);
|
---|
| 49 | }
|
---|
| 50 | }
|
---|
| 51 |
|
---|
| 52 | return @filtered_tags;
|
---|
| 53 | }
|
---|
| 54 |
|
---|
| 55 | sub extract_urls {
|
---|
| 56 | my (@tags) = @_;
|
---|
| 57 |
|
---|
| 58 | my @urls = ();
|
---|
| 59 |
|
---|
| 60 | my $t;
|
---|
| 61 | foreach $t (@tags) {
|
---|
| 62 | if ($t =~ m/href=([^ ]+)/i) {
|
---|
| 63 | my $url = $1;
|
---|
| 64 | $url =~ s/&/&/g;
|
---|
| 65 | push(@urls,$url);
|
---|
| 66 | }
|
---|
| 67 | }
|
---|
| 68 |
|
---|
| 69 | return @urls;
|
---|
| 70 | }
|
---|
| 71 |
|
---|
| 72 | sub get_gi_page
|
---|
| 73 | {
|
---|
| 74 | my ($cgi_base,$cgi_call,$downloadto_fname) = @_;
|
---|
| 75 |
|
---|
| 76 | my $full_url = "$cgi_base$cgi_call";
|
---|
| 77 |
|
---|
| 78 | if ((!-e $downloadto_fname) || (-z $downloadto_fname)) {
|
---|
| 79 | my $cmd = "wget -nv -T 10 -nc -U \"Mozilla\" -O \"$downloadto_fname\" \"$full_url\"";
|
---|
[10112] | 80 | ## print STDERR "*** wget cmd:\n $cmd\n";
|
---|
| 81 |
|
---|
[7492] | 82 | `$cmd`;
|
---|
| 83 | }
|
---|
| 84 |
|
---|
| 85 | if (-z $downloadto_fname) {
|
---|
| 86 | print STDERR "Warning: downloaded file 0 bytes!\n";
|
---|
| 87 | }
|
---|
| 88 | }
|
---|
| 89 |
|
---|
| 90 |
|
---|
| 91 | sub parse_gi_search_page
|
---|
| 92 | {
|
---|
| 93 | my ($ga_base,$search_term_dir,$downloaded_fname,$currpage_url) = @_;
|
---|
| 94 |
|
---|
| 95 | my $nextpage_url = undef;
|
---|
| 96 |
|
---|
| 97 | my @imgref_urls = ();
|
---|
| 98 |
|
---|
| 99 | my $downloaded_text = readin_html($downloaded_fname);
|
---|
| 100 | if (defined $downloaded_text) {
|
---|
| 101 | my @anchor_tags = stripout_anchortags($downloaded_text);
|
---|
| 102 |
|
---|
| 103 | my @thumbimg_tags = filter_tags("imgres\\?",@anchor_tags);
|
---|
| 104 | my @nextpage_tags = filter_tags("images\\?.*?start=\\d+",@anchor_tags);
|
---|
| 105 |
|
---|
| 106 | my @thumbimg_urls = extract_urls(@thumbimg_tags);
|
---|
| 107 | my @nextpage_urls = extract_urls(@nextpage_tags);
|
---|
| 108 |
|
---|
| 109 | my $curr_start = 0;
|
---|
| 110 | if ($currpage_url =~ m/start=(\d+)/) {
|
---|
| 111 | $curr_start = $1;
|
---|
| 112 | }
|
---|
| 113 |
|
---|
| 114 | my $pot_url;
|
---|
| 115 | foreach $pot_url (@nextpage_urls) {
|
---|
| 116 |
|
---|
| 117 | my ($next_start) = ($pot_url =~ m/start=(\d+)/);
|
---|
| 118 | if ($next_start>$curr_start) {
|
---|
| 119 | $nextpage_url = $pot_url;
|
---|
| 120 | last;
|
---|
| 121 | }
|
---|
| 122 | }
|
---|
| 123 |
|
---|
| 124 | # print "-" x 40, "\n";
|
---|
| 125 | my $c = 1;
|
---|
| 126 | my $p = 1;
|
---|
| 127 |
|
---|
| 128 | foreach my $tvu (@thumbimg_urls) {
|
---|
| 129 | my ($img_url) = ($tvu =~ m/imgurl=([^&]*)/);
|
---|
| 130 | $img_url =~ s/%25/%/g;
|
---|
| 131 |
|
---|
| 132 | my ($imgref_url) = ($tvu =~ m/imgrefurl=([^&]*)/);
|
---|
| 133 | ## print STDERR "****imgref_url = $imgref_url\n";
|
---|
| 134 | $imgref_url =~ s/%25/%/g;
|
---|
| 135 |
|
---|
| 136 | my ($img_ext) = ($img_url =~ m/\.(\w+)$/);
|
---|
| 137 | $img_ext = lc($img_ext);
|
---|
| 138 |
|
---|
[8889] | 139 | # remove http:// if there, so later we can explicitly add it in
|
---|
| 140 | $img_url =~ s/^http:\/\///;
|
---|
| 141 |
|
---|
[7492] | 142 | print "Downloading image url http://$img_url\n";
|
---|
| 143 | my $output_fname = "$search_term_dir/img_$c.$img_ext";
|
---|
| 144 |
|
---|
| 145 | get_gi_page("http://",$img_url,$output_fname);
|
---|
| 146 |
|
---|
| 147 | if (-s $output_fname == 0) {
|
---|
| 148 | unlink $output_fname;
|
---|
| 149 | }
|
---|
| 150 | elsif (system("identify \"$output_fname\"") > 0 ) {
|
---|
| 151 | print STDERR "**** NOT JPEG: output_fname \n";
|
---|
| 152 | unlink $output_fname;
|
---|
| 153 | }
|
---|
| 154 | else {
|
---|
| 155 |
|
---|
| 156 | my $command = "identify \"$output_fname\" 2>&1";
|
---|
| 157 | my $result = `$command`;
|
---|
| 158 |
|
---|
| 159 | my $type = 'unknown';
|
---|
| 160 | my $width = 'unknown';
|
---|
| 161 | my $height = 'unknown';
|
---|
| 162 |
|
---|
| 163 | my $image_safe = quotemeta $output_fname;
|
---|
| 164 | if ($result =~ /^$image_safe (\w+) (\d+)x(\d+)/) {
|
---|
| 165 | $type = $1;
|
---|
| 166 | $width = $2;
|
---|
| 167 | $height = $3;
|
---|
| 168 | }
|
---|
| 169 |
|
---|
| 170 | if (($width ne "unknown") && ($height ne "unknown")) {
|
---|
| 171 | if (($width>200) || ($height>200)) {
|
---|
| 172 | `convert \"$output_fname\" -resize 200x200 /tmp/x.jpg`;
|
---|
| 173 | `/bin/mv /tmp/x.jpg \"$output_fname\"`;
|
---|
| 174 | }
|
---|
| 175 | }
|
---|
| 176 | $c++;
|
---|
| 177 | }
|
---|
| 178 |
|
---|
| 179 | push(@imgref_urls,$imgref_url);
|
---|
| 180 |
|
---|
| 181 | last if ($c==3); # Only take first 2
|
---|
| 182 |
|
---|
| 183 | $p++;
|
---|
| 184 |
|
---|
| 185 | if ($p==20) {
|
---|
| 186 | print STDERR "*** Unable to get enough images after 20 passes\n";
|
---|
| 187 | last;
|
---|
| 188 | }
|
---|
| 189 |
|
---|
| 190 |
|
---|
| 191 | }
|
---|
| 192 |
|
---|
| 193 | if (defined $nextpage_url) {
|
---|
| 194 | print "Next page URL:\n";
|
---|
| 195 | print_tags($nextpage_url);
|
---|
| 196 | }
|
---|
| 197 | # print "-" x 40, "\n";
|
---|
| 198 | }
|
---|
| 199 |
|
---|
| 200 | return ($nextpage_url, \@imgref_urls);
|
---|
| 201 | }
|
---|
| 202 |
|
---|
| 203 | sub make_search_term_safe
|
---|
| 204 | {
|
---|
| 205 | my ($search_terms) = @_;
|
---|
| 206 |
|
---|
| 207 | my $search_term_safe = join("+",@$search_terms);
|
---|
| 208 | $search_term_safe =~ s/\"/%22/g;
|
---|
| 209 | $search_term_safe =~ s/ /+/g;
|
---|
| 210 |
|
---|
| 211 | return $search_term_safe;
|
---|
| 212 | }
|
---|
| 213 |
|
---|
| 214 | sub gi_query_url
|
---|
| 215 | {
|
---|
| 216 | my ($search_term) = @_;
|
---|
| 217 |
|
---|
| 218 | my $search_term_safe = make_search_term_safe($search_term);
|
---|
| 219 |
|
---|
| 220 | my $nextpage_url
|
---|
| 221 | = "/images?as_filetype=jpg&imgc=color\&ie=UTF-8\&oe=UTF-8\&hl=en\&btnG=Google+Search";
|
---|
| 222 | $nextpage_url .= "\&q=$search_term_safe";
|
---|
| 223 |
|
---|
| 224 | return $nextpage_url;
|
---|
| 225 | }
|
---|
| 226 |
|
---|
| 227 | sub gi_url_base
|
---|
| 228 | {
|
---|
| 229 | return "http://images.google.com";
|
---|
| 230 | }
|
---|
| 231 |
|
---|
| 232 | sub giget
|
---|
| 233 | {
|
---|
| 234 | my ($search_terms,$output_dir) = @_;
|
---|
| 235 | my $imgref_urls = [];
|
---|
| 236 |
|
---|
| 237 | if (!-e $output_dir) {
|
---|
| 238 | mkdir($output_dir);
|
---|
| 239 |
|
---|
| 240 | }
|
---|
| 241 |
|
---|
[10112] | 242 | print STDERR "Searching Google Images for: ", join(", ",@$search_terms), "\n";
|
---|
[7492] | 243 |
|
---|
| 244 | my $gi_base = gi_url_base();
|
---|
| 245 | my $nextpage_url = gi_query_url($search_terms);
|
---|
| 246 |
|
---|
| 247 | $respage_fname = "$output_dir/respage1.html";
|
---|
| 248 | get_gi_page($gi_base,$nextpage_url,$respage_fname);
|
---|
| 249 |
|
---|
| 250 | ($nextpage_url, $imgref_urls)
|
---|
| 251 | = parse_gi_search_page($gi_base,$output_dir,
|
---|
| 252 | $respage_fname,$nextpage_url);
|
---|
| 253 | # else {
|
---|
| 254 | # print STDERR " Images already mirrored\n";
|
---|
| 255 | # }
|
---|
| 256 |
|
---|
| 257 | print STDERR "-" x 40, "\n";
|
---|
| 258 |
|
---|
| 259 | return $imgref_urls;
|
---|
| 260 | }
|
---|
| 261 |
|
---|
| 262 |
|
---|
| 263 | 1;
|
---|