--- test/html-webhacc/cc.cgi	2007/11/04 09:15:02	1.22
+++ test/html-webhacc/cc.cgi	2008/02/09 12:22:19	1.30
@@ -1,5 +1,6 @@
 #!/usr/bin/perl
 use strict;
+use utf8;
 
 use lib qw[/home/httpd/html/www/markup/html/whatpm
            /home/wakaba/work/manakai2/lib];
@@ -51,7 +52,6 @@
 
   $| = 0;
   my $input = get_input_document ($http, $dom);
-  my $inner_html_element = $http->get_parameter ('e');
   my $char_length = 0;
   my %time;
 
@@ -61,7 +61,12 @@
 <dt>Request URI</dt>
     <dd><code class="URI" lang="">&lt;<a href="@{[htescape $input->{request_uri}]}">@{[htescape $input->{request_uri}]}</a>&gt;</code></dd>
 <dt>Document URI</dt>
-    <dd><code class="URI" lang="">&lt;<a href="@{[htescape $input->{uri}]}">@{[htescape $input->{uri}]}</a>&gt;</code></dd>
+    <dd><code class="URI" lang="">&lt;<a href="@{[htescape $input->{uri}]}" id=anchor-document-uri>@{[htescape $input->{uri}]}</a>&gt;</code>
+    <script>
+      document.title = '<'
+          + document.getElementById ('anchor-document-uri').href + '> \\u2014 '
+          + document.title;
+    </script></dd>
 ]; # no </dl> yet
   push @nav, ['#document-info' => 'Information'];
 
@@ -73,7 +78,7 @@
     <dd><code class="URI" lang="">&lt;<a href="@{[htescape $input->{base_uri}]}">@{[htescape $input->{base_uri}]}</a>&gt;</code></dd>
 <dt>Internet Media Type</dt>
     <dd><code class="MIME" lang="en">@{[htescape $input->{media_type}]}</code>
-    @{[$input->{media_type_overridden} ? '<em>(overridden)</em>' : '']}</dd>
+    @{[$input->{media_type_overridden} ? '<em>(overridden)</em>' : defined $input->{official_type} ? $input->{media_type} eq $input->{official_type} ? '' : '<em>(sniffed; official type is: <code class=MIME lang=en>'.htescape ($input->{official_type}).'</code>)' : '<em>(sniffed)</em>']}</dd>
 <dt>Character Encoding</dt>
     <dd>@{[defined $input->{charset} ? '<code class="charset" lang="en">'.htescape ($input->{charset}).'</code>' : '(none)']}
     @{[$input->{charset_overridden} ? '<em>(overridden)</em>' : '']}</dd>
@@ -92,7 +97,8 @@
 
   if ($input->{media_type} eq 'text/html') {
     ($doc, $el) = print_syntax_error_html_section ($input, $result);
-    print_source_string_section (\($input->{s}), $input->{charset});
+    print_source_string_section
+        (\($input->{s}), $input->{charset} || $doc->input_encoding);
   } elsif ({
             'text/xml' => 1,
             'application/atom+xml' => 1,
@@ -109,7 +115,7 @@
     print_source_string_section (\($input->{s}), 'utf-8');
   } else {
     ## TODO: Change HTTP status code??
-    print_result_unknown_type_section ($input);
+    print_result_unknown_type_section ($input, $result);
   }
 
   if (defined $doc or defined $el) {
@@ -142,7 +148,8 @@
 </html>
 ];
 
-  for (qw/decode parse parse_xml parse_manifest check check_manifest/) {
+  for (qw/decode parse parse_html parse_xml parse_manifest
+          check check_manifest/) {
     next unless defined $time{$_};
     open my $file, '>>', ".cc-$_.txt" or die ".cc-$_.txt: $!";
     print $file $char_length, "\t", $time{$_}, "\n";
@@ -159,7 +166,7 @@
       $result->{conforming_min} = 0;
     } elsif ($err->{level} eq 'w' or $err->{level} eq 'g') {
       $result->{$layer}->{warning}++;
-    } elsif ($err->{level} eq 'unsupported') {
+    } elsif ($err->{level} eq 'u' or $err->{level} eq 'unsupported') {
       $result->{$layer}->{unsupported}++;
       $result->{unsupported} = 1;
     } else {
@@ -217,13 +224,7 @@
   
   require Encode;
   require Whatpm::HTML;
-
-  $input->{charset} ||= 'ISO-8859-1'; ## TODO: for now.
   
-  my $time1 = time;
-  my $t = Encode::decode ($input->{charset}, $input->{s});
-  $time{decode} = time - $time1;
-
   print STDOUT qq[
 <div id="parse-errors" class="section">
 <h2>Parse Errors</h2>
@@ -243,23 +244,35 @@
     $type =~ tr/ /-/;
     $type =~ s/\|/%7C/g;
     $msg .= qq[ [<a href="../error-description#@{[htescape ($type)]}">Description</a>]];
-    print STDOUT qq[<dd class="$cls">$msg</dd>\n];
+    print STDOUT qq[<dd class="$cls">], get_error_level_label (\%opt);
+    print STDOUT qq[$msg</dd>\n];
 
     add_error ('syntax', \%opt => $result);
   };
 
   my $doc = $dom->create_document;
   my $el;
-  $time1 = time;
+  my $inner_html_element = $http->get_parameter ('e');
   if (defined $inner_html_element and length $inner_html_element) {
+    $input->{charset} ||= 'windows-1252'; ## TODO: for now.
+    my $time1 = time;
+    my $t = Encode::decode ($input->{charset}, $input->{s});
+    $time{decode} = time - $time1;
+    
     $el = $doc->create_element_ns
         ('http://www.w3.org/1999/xhtml', [undef, $inner_html_element]);
+    $time1 = time;
     Whatpm::HTML->set_inner_html ($el, $t, $onerror);
+    $time{parse} = time - $time1;
   } else {
-    Whatpm::HTML->parse_string ($t => $doc, $onerror);
+    my $time1 = time;
+    Whatpm::HTML->parse_byte_string
+        ($input->{charset}, $input->{s} => $doc, $onerror);
+    $time{parse_html} = time - $time1;
   }
-  $time{parse} = time - $time1;
-
+  $doc->manakai_charset ($input->{official_charset})
+      if defined $input->{official_charset};
+  
   print STDOUT qq[</dl></div>];
 
   return ($doc, $el);
@@ -299,6 +312,8 @@
   my $doc = Message::DOM::XMLParserTemp->parse_byte_stream
       ($fh => $dom, $onerror, charset => $input->{charset});
   $time{parse_xml} = time - $time1;
+  $doc->manakai_charset ($input->{official_charset})
+      if defined $input->{official_charset};
 
   print STDOUT qq[</dl></div>];
 
@@ -324,7 +339,8 @@
     $type =~ tr/ /-/;
     $type =~ s/\|/%7C/g;
     $msg .= qq[ [<a href="../error-description#@{[htescape ($type)]}">Description</a>]];
-    print STDOUT qq[<dd class="$cls">$msg</dd>\n];
+    print STDOUT qq[<dd class="$cls">], get_error_level_label (\%opt);
+    print STDOUT qq[$msg</dd>\n];
 
     add_error ('syntax', \%opt => $result);
   };
@@ -387,7 +403,7 @@
         $r .= '<ul class="attributes">';
         for my $attr (sort {$a->[0] cmp $b->[0]} map { [$_->name, $_->value, $_->namespace_uri, 'node-'.refaddr $_] }
                       @{$child->attributes}) {
-          $r .= qq[<li id="$attr->[3]" class="tree-attribute"><code title="@{[defined $_->[2] ? $_->[2] : '']}">] . htescape ($attr->[0]) . '</code> = '; ## ISSUE: case?
+          $r .= qq[<li id="$attr->[3]" class="tree-attribute"><code title="@{[defined $attr->[2] ? htescape ($attr->[2]) : '']}">] . htescape ($attr->[0]) . '</code> = '; ## ISSUE: case?
           $r .= '<q>' . htescape ($attr->[1]) . '</q></li>'; ## TODO: children
         }
         $r .= '</ul>';
@@ -408,6 +424,21 @@
     } elsif ($nt == $child->DOCUMENT_NODE) {
       $r .= qq'<li id="$node_id" class="tree-document">Document';
       $r .= qq[<ul class="attributes">];
+      my $cp = $child->manakai_charset;
+      if (defined $cp) {
+        $r .= qq[<li><code>charset</code> parameter = <code>];
+        $r .= htescape ($cp) . qq[</code></li>];
+      }
+      $r .= qq[<li><code>inputEncoding</code> = ];
+      my $ie = $child->input_encoding;
+      if (defined $ie) {
+        $r .= qq[<code>@{[htescape ($ie)]}</code>];
+        if ($child->manakai_has_bom) {
+          $r .= qq[ (with <code class=charname><abbr>BOM</abbr></code>)];
+        }
+      } else {
+        $r .= qq[(<code>null</code>)];
+      }
       $r .= qq[<li>@{[scalar get_text ('manakaiIsHTML:'.($child->manakai_is_html?1:0))]}</li>];
       $r .= qq[<li>@{[scalar get_text ('manakaiCompatMode:'.$child->manakai_compat_mode)]}</li>];
       unless ($child->manakai_is_html) {
@@ -506,7 +537,8 @@
     $type =~ s/\|/%7C/g;
     $msg .= qq[ [<a href="../error-description#@{[htescape ($type)]}">Description</a>]];
     print STDOUT qq[<dt class="$cls">] . get_error_label (\%opt) .
-        qq[</dt>\n<dd class="$cls">], $msg, "</dd>\n";
+        qq[</dt>\n<dd class="$cls">], get_error_level_label (\%opt);
+    print STDOUT $msg, "</dd>\n";
     add_error ('structure', \%opt => $result);
   };
 
@@ -691,10 +723,13 @@
   print STDOUT qq[<table>
 <colgroup><col><colgroup><col><col><col><colgroup><col>
 <thead>
-<tr><th scope=col></th><th scope=col><em class=rfc2119>MUST</em>-level
-Errors</th><th scope=col><em class=rfc2119>SHOULD</em>-level
-Errors</th><th scope=col>Warnings</th><th scope=col>Score</th></tr>
-</thead><tbody>];
+<tr><th scope=col></th>
+<th scope=col><a href="../error-description#level-m"><em class=rfc2119>MUST</em>‐level
+Errors</a></th>
+<th scope=col><a href="../error-description#level-s"><em class=rfc2119>SHOULD</em>‐level
+Errors</a></th>
+<th scope=col><a href="../error-description#level-w">Warnings</a></th>
+<th scope=col>Score</th></tr></thead><tbody>];
 
   my $must_error = 0;
   my $should_error = 0;
@@ -752,15 +787,28 @@
   push @nav, ['#result-summary' => 'Result'];
 } # print_result_section
 
-sub print_result_unknown_type_section ($) {
-  my $input = shift;
+sub print_result_unknown_type_section ($$) {
+  my ($input, $result) = @_;
 
+  my $euri = htescape ($input->{uri});
   print STDOUT qq[
-<div id="result-summary" class="section">
-<p><em>Media type <code class="MIME" lang="en">@{[htescape $input->{media_type}]}</code> is not supported!</em></p>
+<div id="parse-errors" class="section">
+<h2>Errors</h2>
+
+<dl>
+<dt class=unsupported><code>&lt;<a href="$euri">$euri</a>&gt;</code></dt>
+    <dd class=unsupported><strong><a href="../error-description#level-u">Not
+        supported</a></strong>:
+    Media type
+    <code class="MIME" lang="en">@{[htescape $input->{media_type}]}</code>
+    is not supported.</dd>
+</dl>
 </div>
 ];
-  push @nav, ['#result-summary' => 'Result'];
+  push @nav, ['#parse-errors' => 'Errors'];
+  add_error (char => {level => 'u'} => $result);
+  add_error (syntax => {level => 'u'} => $result);
+  add_error (structure => {level => 'u'} => $result);
 } # print_result_unknown_type_section
 
 sub print_result_input_error_section ($) {
@@ -787,7 +835,7 @@
 
   if (defined $err->{node}) {
     $r .= ' ' if length $r;
-    $r = get_node_path ($err->{node});
+    $r = get_node_link ($err->{node});
   }
 
   if (defined $err->{index}) {
@@ -803,6 +851,32 @@
   return $r;
 } # get_error_label
 
+sub get_error_level_label ($) {
+  my $err = shift;
+
+  my $r = '';
+
+  if (not defined $err->{level} or $err->{level} eq 'm') {
+    $r = qq[<strong><a href="../error-description#level-m"><em class=rfc2119>MUST</em>‐level
+        error</a></strong>: ];
+  } elsif ($err->{level} eq 's') {
+    $r = qq[<strong><a href="../error-description#level-s"><em class=rfc2119>SHOULD</em>‐level
+        error</a></strong>: ];
+  } elsif ($err->{level} eq 'w') {
+    $r = qq[<strong><a href="../error-description#level-w">Warning</a></strong>:
+        ];
+  } elsif ($err->{level} eq 'u' or $err->{level} eq 'unsupported') {
+    $r = qq[<strong><a href="../error-description#level-u">Not
+        supported</a></strong>: ];
+  } else {
+    my $elevel = htescape ($err->{level});
+    $r = qq[<strong><a href="../error-description#level-$elevel">$elevel</a></strong>:
+        ];
+  }
+
+  return $r;
+} # get_error_level_label
+
 sub get_node_path ($) {
   my $node = shift;
   my @r;
@@ -840,7 +914,8 @@
 
 sub load_text_catalog ($) {
   my $lang = shift; # MUST be a canonical lang name
-  open my $file, '<', "cc-msg.$lang.txt" or die "$0: cc-msg.$lang.txt: $!";
+  open my $file, '<:utf8', "cc-msg.$lang.txt"
+      or die "$0: cc-msg.$lang.txt: $!";
   while (<$file>) {
     if (s/^([^;]+);([^;]*);//) {
       my ($type, $cls, $msg) = ($1, $2, $_);
@@ -853,6 +928,7 @@
 sub get_text ($) {
   my ($type, $level, $node) = @_;
   $type = $level . ':' . $type if defined $level;
+  $level = 'm' unless defined $level;
   my @arg;
   {
     if (defined $Msg->{$type}) {
@@ -877,13 +953,13 @@
           ? htescape ($node->owner_element->manakai_local_name)
           : ''
       }ge;
-      return ($type, $Msg->{$type}->[0], $msg);
+      return ($type, 'level-' . $level . ' ' . $Msg->{$type}->[0], $msg);
     } elsif ($type =~ s/:([^:]*)$//) {
       unshift @arg, $1;
       redo;
     }
   }
-  return ($type, '', htescape ($_[0]));
+  return ($type, 'level-'.$level, htescape ($_[0]));
 } # get_text
 
 }
@@ -939,6 +1015,7 @@
     $ua->protocols_allowed ([qw/http/]);
     $ua->max_size (1000_000);
     my $req = HTTP::Request->new (GET => $request_uri);
+    $req->header ('Accept-Encoding' => 'identity, *; q=0');
     my $res = $ua->request ($req);
     ## TODO: 401 sets |is_success| true.
     if ($res->is_success or $http->get_parameter ('error-page')) {
@@ -948,12 +1025,10 @@
 
       ## TODO: More strict parsing...
       my $ct = $res->header ('Content-Type');
-      if (defined $ct and $ct =~ m#^([0-9A-Za-z._+-]+/[0-9A-Za-z._+-]+)#) {
-        $r->{media_type} = lc $1;
-      }
       if (defined $ct and $ct =~ /;\s*charset\s*=\s*"?([^\s;"]+)"?/i) {
         $r->{charset} = lc $1;
         $r->{charset} =~ tr/\\//d;
+        $r->{official_charset} = $r->{charset};
       }
 
       my $input_charset = $http->get_parameter ('charset');
@@ -961,9 +1036,22 @@
         $r->{charset_overridden}
             = (not defined $r->{charset} or $r->{charset} ne $input_charset);
         $r->{charset} = $input_charset;
-      } 
+      }
+
+      ## TODO: Support for HTTP Content-Encoding
 
       $r->{s} = ''.$res->content;
+
+      require Whatpm::ContentType;
+      ($r->{official_type}, $r->{media_type})
+          = Whatpm::ContentType->get_sniffed_type
+              (get_file_head => sub {
+                 return substr $r->{s}, 0, shift;
+               },
+               http_content_type_byte => $ct,
+               has_http_content_encoding =>
+                   defined $res->header ('Content-Encoding'),
+               supported_image_types => {});
     } else {
       $r->{uri} = $res->request->uri;
       $r->{request_uri} = $request_uri;
@@ -984,7 +1072,18 @@
     $r->{charset} = ''.$http->get_parameter ('_charset_');
     $r->{charset} =~ s/\s+//g;
     $r->{charset} = 'utf-8' if $r->{charset} eq '';
+    $r->{official_charset} = $r->{charset};
     $r->{header_field} = [];
+
+    require Whatpm::ContentType;
+    ($r->{official_type}, $r->{media_type})
+        = Whatpm::ContentType->get_sniffed_type
+            (get_file_head => sub {
+               return substr $r->{s}, 0, shift;
+             },
+             http_content_type_byte => undef,
+             has_http_content_encoding => 0,
+             supported_image_types => {});
   }
 
   my $input_format = $http->get_parameter ('i');
@@ -1001,6 +1100,7 @@
   if ($r->{media_type} eq 'text/xml') {
     unless (defined $r->{charset}) {
       $r->{charset} = 'us-ascii';
+      $r->{official_charset} = $r->{charset};
     } elsif ($r->{charset_overridden} and $r->{charset} eq 'us-ascii') {
       $r->{charset_overridden} = 0;
     }
@@ -1051,4 +1151,4 @@
 
 =cut
 
-## $Date: 2007/11/04 09:15:02 $
+## $Date: 2008/02/09 12:22:19 $