Whatpm/HTML/Tokenizer.pm.src

package Whatpm::HTML::Tokenizer;
use strict;
our $VERSION=do{my @r=(q$Revision: 1.14 $=~/\d+/g);sprintf "%d."."%02d" x $#r,@r};

BEGIN {
  require Exporter;
  push our @ISA, 'Exporter';

  our @EXPORT_OK = qw(
    DOCTYPE_TOKEN
    COMMENT_TOKEN
    START_TAG_TOKEN
    END_TAG_TOKEN
    END_OF_FILE_TOKEN
    CHARACTER_TOKEN
    PI_TOKEN
    ABORT_TOKEN
    END_OF_DOCTYPE_TOKEN
    ATTLIST_TOKEN
    ELEMENT_TOKEN
    GENERAL_ENTITY_TOKEN
    PARAMETER_ENTITY_TOKEN
    NOTATION_TOKEN
  );
  
  our %EXPORT_TAGS = (
    token => [qw(
      DOCTYPE_TOKEN
      COMMENT_TOKEN
      START_TAG_TOKEN
      END_TAG_TOKEN
      END_OF_FILE_TOKEN
      CHARACTER_TOKEN
      PI_TOKEN
      ABORT_TOKEN
      END_OF_DOCTYPE_TOKEN
      ATTLIST_TOKEN
      ELEMENT_TOKEN
      GENERAL_ENTITY_TOKEN
      PARAMETER_ENTITY_TOKEN
      NOTATION_TOKEN
    )],
  );
}

## NOTE: Differences from the XML5 draft are marked as "XML5:".

## Token types

sub DOCTYPE_TOKEN () { 1 } ## XML5: No DOCTYPE token.
sub COMMENT_TOKEN () { 2 }
sub START_TAG_TOKEN () { 3 }
sub END_TAG_TOKEN () { 4 }
sub END_OF_FILE_TOKEN () { 5 }
sub CHARACTER_TOKEN () { 6 }
sub PI_TOKEN () { 7 } ## NOTE: XML only.
sub ABORT_TOKEN () { 8 } ## NOTE: For internal processing.
sub END_OF_DOCTYPE_TOKEN () { 9 } ## NOTE: XML only.
sub ATTLIST_TOKEN () { 10 } ## NOTE: XML only.
sub ELEMENT_TOKEN () { 11 } ## NOTE: XML only.
sub GENERAL_ENTITY_TOKEN () { 12 } ## NOTE: XML only.
sub PARAMETER_ENTITY_TOKEN () { 13 } ## NOTE: XML only.
sub NOTATION_TOKEN () { 14 } ## NOTE: XML only.

## XML5: XML5 has "empty tag token".  In this implementation, it is
## represented as a start tag token with $self->{self_closing} flag
## set to true.

## XML5: XML5 has "short end tag token".  In this implementation, it
## is represented as an end tag token with $token->{tag_name} flag set
## to an empty string.

package Whatpm::HTML;

BEGIN { Whatpm::HTML::Tokenizer->import (':token') }

## Content model flags

sub CM_ENTITY () { 0b001 } # & markup in data
sub CM_LIMITED_MARKUP () { 0b010 } # < markup in data (limited)
sub CM_FULL_MARKUP () { 0b100 } # < markup in data (any)

sub PLAINTEXT_CONTENT_MODEL () { 0 }
sub CDATA_CONTENT_MODEL () { CM_LIMITED_MARKUP }
sub RCDATA_CONTENT_MODEL () { CM_ENTITY | CM_LIMITED_MARKUP }
sub PCDATA_CONTENT_MODEL () { CM_ENTITY | CM_FULL_MARKUP }

## Tokenizer states

sub DATA_STATE () { 0 }
#sub ENTITY_DATA_STATE () { 1 }
sub TAG_OPEN_STATE () { 2 }
sub CLOSE_TAG_OPEN_STATE () { 3 }
sub TAG_NAME_STATE () { 4 }
sub BEFORE_ATTRIBUTE_NAME_STATE () { 5 }
sub ATTRIBUTE_NAME_STATE () { 6 }
sub AFTER_ATTRIBUTE_NAME_STATE () { 7 }
sub BEFORE_ATTRIBUTE_VALUE_STATE () { 8 }
sub ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE () { 9 }
sub ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE () { 10 }
sub ATTRIBUTE_VALUE_UNQUOTED_STATE () { 11 }
#sub ENTITY_IN_ATTRIBUTE_VALUE_STATE () { 12 }
sub MARKUP_DECLARATION_OPEN_STATE () { 13 }
sub COMMENT_START_STATE () { 14 }
sub COMMENT_START_DASH_STATE () { 15 }
sub COMMENT_STATE () { 16 }
sub COMMENT_END_STATE () { 17 }
sub COMMENT_END_DASH_STATE () { 18 }
sub BOGUS_COMMENT_STATE () { 19 }
sub DOCTYPE_STATE () { 20 }
sub BEFORE_DOCTYPE_NAME_STATE () { 21 }
sub DOCTYPE_NAME_STATE () { 22 }
sub AFTER_DOCTYPE_NAME_STATE () { 23 }
sub BEFORE_DOCTYPE_PUBLIC_IDENTIFIER_STATE () { 24 }
sub DOCTYPE_PUBLIC_IDENTIFIER_DOUBLE_QUOTED_STATE () { 25 }
sub DOCTYPE_PUBLIC_IDENTIFIER_SINGLE_QUOTED_STATE () { 26 }
sub AFTER_DOCTYPE_PUBLIC_IDENTIFIER_STATE () { 27 }
sub BEFORE_DOCTYPE_SYSTEM_IDENTIFIER_STATE () { 28 }
sub DOCTYPE_SYSTEM_IDENTIFIER_DOUBLE_QUOTED_STATE () { 29 }
sub DOCTYPE_SYSTEM_IDENTIFIER_SINGLE_QUOTED_STATE () { 30 }
sub AFTER_DOCTYPE_SYSTEM_IDENTIFIER_STATE () { 31 }
sub BOGUS_DOCTYPE_STATE () { 32 }
sub AFTER_ATTRIBUTE_VALUE_QUOTED_STATE () { 33 }
sub SELF_CLOSING_START_TAG_STATE () { 34 }
sub CDATA_SECTION_STATE () { 35 }
sub MD_HYPHEN_STATE () { 36 } # "markup declaration open state" in the spec
sub MD_DOCTYPE_STATE () { 37 } # "markup declaration open state" in the spec
sub MD_CDATA_STATE () { 38 } # "markup declaration open state" in the spec
sub CDATA_RCDATA_CLOSE_TAG_STATE () { 39 } # "close tag open state" in the spec
sub CDATA_SECTION_MSE1_STATE () { 40 } # "CDATA section state" in the spec
sub CDATA_SECTION_MSE2_STATE () { 41 } # "CDATA section state" in the spec
sub PUBLIC_STATE () { 42 } # "after DOCTYPE name state" in the spec
sub SYSTEM_STATE () { 43 } # "after DOCTYPE name state" in the spec
## NOTE: "Entity data state", "entity in attribute value state", and
## "consume a character reference" algorithm are jointly implemented
## using the following six states:
sub ENTITY_STATE () { 44 }
sub ENTITY_HASH_STATE () { 45 }
sub NCR_NUM_STATE () { 46 }
sub HEXREF_X_STATE () { 47 }
sub HEXREF_HEX_STATE () { 48 }
sub ENTITY_NAME_STATE () { 49 }
sub PCDATA_STATE () { 50 } # "data state" in the spec

## XML-only states
sub PI_STATE () { 51 }
sub PI_TARGET_STATE () { 52 }
sub PI_TARGET_AFTER_STATE () { 53 }
sub PI_DATA_STATE () { 54 }
sub PI_AFTER_STATE () { 55 }
sub PI_DATA_AFTER_STATE () { 56 }
sub DOCTYPE_INTERNAL_SUBSET_STATE () { 57 }
sub DOCTYPE_INTERNAL_SUBSET_AFTER_STATE () { 58 }
sub BOGUS_DOCTYPE_INTERNAL_SUBSET_AFTER_STATE () { 59 }
sub DOCTYPE_TAG_STATE () { 60 }
sub DOCTYPE_MARKUP_DECLARATION_OPEN_STATE () { 61 }
sub MD_ATTLIST_STATE () { 62 }
sub MD_E_STATE () { 63 }
sub MD_ELEMENT_STATE () { 64 }
sub MD_ENTITY_STATE () { 65 }
sub MD_NOTATION_STATE () { 66 }
sub DOCTYPE_MD_STATE () { 67 }
sub BEFORE_MD_NAME_STATE () { 68 }
sub MD_NAME_STATE () { 69 }
sub DOCTYPE_ENTITY_PARAMETER_BEFORE_STATE () { 70 }
sub DOCTYPE_ATTLIST_NAME_AFTER_STATE () { 71 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_NAME_STATE () { 72 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_NAME_AFTER_STATE () { 73 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_STATE () { 74 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_AFTER_STATE () { 75 }
sub BEFORE_ALLOWED_TOKEN_STATE () { 76 }
sub ALLOWED_TOKEN_STATE () { 77 }
sub AFTER_ALLOWED_TOKEN_STATE () { 78 }
sub AFTER_ALLOWED_TOKENS_STATE () { 79 }
sub BEFORE_ATTR_DEFAULT_STATE () { 80 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE () { 81 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_STATE () { 82 }
sub DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_AFTER_STATE () { 83 }
sub AFTER_ATTLIST_ATTR_VALUE_QUOTED_STATE () { 84 }

## Tree constructor state constants (see Whatpm::HTML for the full
## list and descriptions)

sub IN_FOREIGN_CONTENT_IM () { 0b100000000000 }
sub FOREIGN_EL () { 0b1_00000000000 }

## Character reference mappings

my $charref_map = {
  0x0D => 0x000A,
  0x80 => 0x20AC,
  0x81 => 0xFFFD,
  0x82 => 0x201A,
  0x83 => 0x0192,
  0x84 => 0x201E,
  0x85 => 0x2026,
  0x86 => 0x2020,
  0x87 => 0x2021,
  0x88 => 0x02C6,
  0x89 => 0x2030,
  0x8A => 0x0160,
  0x8B => 0x2039,
  0x8C => 0x0152,
  0x8D => 0xFFFD,
  0x8E => 0x017D,
  0x8F => 0xFFFD,
  0x90 => 0xFFFD,
  0x91 => 0x2018,
  0x92 => 0x2019,
  0x93 => 0x201C,
  0x94 => 0x201D,
  0x95 => 0x2022,
  0x96 => 0x2013,
  0x97 => 0x2014,
  0x98 => 0x02DC,
  0x99 => 0x2122,
  0x9A => 0x0161,
  0x9B => 0x203A,
  0x9C => 0x0153,
  0x9D => 0xFFFD,
  0x9E => 0x017E,
  0x9F => 0x0178,
}; # $charref_map
$charref_map->{$_} = 0xFFFD
    for 0x0000..0x0008, 0x000B, 0x000E..0x001F, 0x007F, 
        0xD800..0xDFFF, 0xFDD0..0xFDDF, ## ISSUE: 0xFDEF
        0xFFFE, 0xFFFF, 0x1FFFE, 0x1FFFF, 0x2FFFE, 0x2FFFF, 0x3FFFE, 0x3FFFF,
        0x4FFFE, 0x4FFFF, 0x5FFFE, 0x5FFFF, 0x6FFFE, 0x6FFFF, 0x7FFFE,
        0x7FFFF, 0x8FFFE, 0x8FFFF, 0x9FFFE, 0x9FFFF, 0xAFFFE, 0xAFFFF,
        0xBFFFE, 0xBFFFF, 0xCFFFE, 0xCFFFF, 0xDFFFE, 0xDFFFF, 0xEFFFE,
        0xEFFFF, 0xFFFFE, 0xFFFFF, 0x10FFFE, 0x10FFFF;

## Implementations MUST act as if state machine in the spec

sub _initialize_tokenizer ($) {
  my $self = shift;

  ## NOTE: Fields set by |new| constructor:
  #$self->{level}
  #$self->{set_nc}
  #$self->{parse_error}
  #$self->{is_xml} (if XML)

  $self->{state} = DATA_STATE; # MUST
  $self->{s_kwd} = ''; # Data state keyword
  #$self->{kwd} = ''; # State-dependent keyword; initialized when used
  #$self->{entity__value}; # initialized when used
  #$self->{entity__match}; # initialized when used
  $self->{content_model} = PCDATA_CONTENT_MODEL; # be
  undef $self->{ct}; # current token
  undef $self->{ca}; # current attribute
  undef $self->{last_stag_name}; # last emitted start tag name
  #$self->{prev_state}; # initialized when used
  delete $self->{self_closing};
  $self->{char_buffer} = '';
  $self->{char_buffer_pos} = 0;
  $self->{nc} = -1; # next input character
  #$self->{next_nc}
  !!!next-input-character;
  $self->{token} = [];
  # $self->{escape}
} # _initialize_tokenizer

## A token has:
##   ->{type} == DOCTYPE_TOKEN, START_TAG_TOKEN, END_TAG_TOKEN, COMMENT_TOKEN,
##       CHARACTER_TOKEN, END_OF_FILE_TOKEN, PI_TOKEN, or ABORT_TOKEN
##   ->{name} (DOCTYPE_TOKEN)
##   ->{tag_name} (START_TAG_TOKEN, END_TAG_TOKEN)
##   ->{target} (PI_TOKEN)
##   ->{pubid} (DOCTYPE_TOKEN)
##   ->{sysid} (DOCTYPE_TOKEN)
##   ->{quirks} == 1 or 0 (DOCTYPE_TOKEN): "force-quirks" flag
##   ->{attributes} isa HASH (START_TAG_TOKEN, END_TAG_TOKEN)
##        ->{name}
##        ->{value}
##        ->{has_reference} == 1 or 0
##        ->{index}: Index of the attribute in a tag.
##   ->{data} (COMMENT_TOKEN, CHARACTER_TOKEN, PI_TOKEN)
##   ->{has_reference} == 1 or 0 (CHARACTER_TOKEN)
##   ->{last_index} (ELEMENT_TOKEN): Next attribute's index - 1.
##   ->{has_internal_subset} = 1 or 0 (DOCTYPE_TOKEN)

## NOTE: The "self-closing flag" is hold as |$self->{self_closing}|.
##     |->{self_closing}| is used to save the value of |$self->{self_closing}|
##     while the token is pushed back to the stack.

## Emitted token MUST immediately be handled by the tree construction state.

## Before each step, UA MAY check to see if either one of the scripts in
## "list of scripts that will execute as soon as possible" or the first
## script in the "list of scripts that will execute asynchronously",
## has completed loading.  If one has, then it MUST be executed
## and removed from the list.

## TODO: Polytheistic slash SHOULD NOT be used. (Applied only to atheists.)
## (This requirement was dropped from HTML5 spec, unfortunately.)

my $is_space = {
  0x0009 => 1, # CHARACTER TABULATION (HT)
  0x000A => 1, # LINE FEED (LF)
  #0x000B => 0, # LINE TABULATION (VT)
  0x000C => 1, # FORM FEED (FF) ## XML5: Not a space character.
  #0x000D => 1, # CARRIAGE RETURN (CR)
  0x0020 => 1, # SPACE (SP)
};

sub _get_next_token ($) {
  my $self = shift;

  if ($self->{self_closing}) {
    !!!parse-error (type => 'nestc', token => $self->{ct});
    ## NOTE: The |self_closing| flag is only set by start tag token.
    ## In addition, when a start tag token is emitted, it is always set to
    ## |ct|.
    delete $self->{self_closing};
  }

  if (@{$self->{token}}) {
    $self->{self_closing} = $self->{token}->[0]->{self_closing};
    return shift @{$self->{token}};
  }

  A: {
    if ($self->{state} == PCDATA_STATE) {
      ## NOTE: Same as |DATA_STATE|, but only for |PCDATA| content model.

      if ($self->{nc} == 0x0026) { # &
        !!!cp (0.1);
        ## NOTE: In the spec, the tokenizer is switched to the 
        ## "entity data state".  In this implementation, the tokenizer
        ## is switched to the |ENTITY_STATE|, which is an implementation
        ## of the "consume a character reference" algorithm.
        $self->{entity_add} = -1;
        $self->{prev_state} = DATA_STATE;
        $self->{state} = ENTITY_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003C) { # <
        !!!cp (0.2);
        $self->{state} = TAG_OPEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (0.3);
        !!!emit ({type => END_OF_FILE_TOKEN,
                  line => $self->{line}, column => $self->{column}});
        last A; ## TODO: ok?
      } else {
        !!!cp (0.4);
        #
      }

      # Anything else
      my $token = {type => CHARACTER_TOKEN,
                   data => chr $self->{nc},
                   line => $self->{line}, column => $self->{column},
                  };
      $self->{read_until}->($token->{data}, q[<&], length $token->{data});

      ## Stay in the state.
      !!!next-input-character;
      !!!emit ($token);
      redo A;
    } elsif ($self->{state} == DATA_STATE) {
      $self->{s_kwd} = '' unless defined $self->{s_kwd};
      if ($self->{nc} == 0x0026) { # &
        $self->{s_kwd} = '';
        if ($self->{content_model} & CM_ENTITY and # PCDATA | RCDATA
            not $self->{escape}) {
          !!!cp (1);
          ## NOTE: In the spec, the tokenizer is switched to the 
          ## "entity data state".  In this implementation, the tokenizer
          ## is switched to the |ENTITY_STATE|, which is an implementation
          ## of the "consume a character reference" algorithm.
          $self->{entity_add} = -1;
          $self->{prev_state} = DATA_STATE;
          $self->{state} = ENTITY_STATE;
          !!!next-input-character;
          redo A;
        } else {
          !!!cp (2);
          #
        }
      } elsif ($self->{nc} == 0x002D) { # -
        if ($self->{content_model} & CM_LIMITED_MARKUP) { # RCDATA | CDATA
          if ($self->{s_kwd} eq '<!-') {
            !!!cp (3);
            $self->{escape} = 1; # unless $self->{escape};
            $self->{s_kwd} = '--';
            #
          } elsif ($self->{s_kwd} eq '-') {
            !!!cp (4);
            $self->{s_kwd} = '--';
            #
          } elsif ($self->{s_kwd} eq '<!' or $self->{s_kwd} eq '-') {
            !!!cp (4.1);
            $self->{s_kwd} .= '-';
            #
          } else {
            !!!cp (5);
            $self->{s_kwd} = '-';
            #
          }
        }
        
        #
      } elsif ($self->{nc} == 0x0021) { # !
        if (length $self->{s_kwd}) {
          !!!cp (5.1);
          $self->{s_kwd} .= '!';
          #
        } else {
          !!!cp (5.2);
          #$self->{s_kwd} = '';
          #
        }
        #
      } elsif ($self->{nc} == 0x003C) { # <
        if ($self->{content_model} & CM_FULL_MARKUP or # PCDATA
            (($self->{content_model} & CM_LIMITED_MARKUP) and # CDATA | RCDATA
             not $self->{escape})) {
          !!!cp (6);
          $self->{state} = TAG_OPEN_STATE;
          !!!next-input-character;
          redo A;
        } else {
          !!!cp (7);
          $self->{s_kwd} = '';
          #
        }
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{escape} and
            ($self->{content_model} & CM_LIMITED_MARKUP)) { # RCDATA | CDATA
          if ($self->{s_kwd} eq '--') {
            !!!cp (8);
            delete $self->{escape};
            #
          } else {
            !!!cp (9);
            #
          }
        } elsif ($self->{is_xml} and $self->{s_kwd} eq ']]') {
          !!!cp (9.1);
          !!!parse-error (type => 'unmatched mse', ## TODO: type
                          line => $self->{line_prev},
                          column => $self->{column_prev} - 1);
          #
        } else {
          !!!cp (10);
          #
        }
        
        $self->{s_kwd} = '';
        #
      } elsif ($self->{nc} == 0x005D) { # ]
        if ($self->{s_kwd} eq ']' or $self->{s_kwd} eq '') {
          !!!cp (10.1);
          $self->{s_kwd} .= ']';
        } elsif ($self->{s_kwd} eq ']]') {
          !!!cp (10.2);
          #
        } else {
          !!!cp (10.3);
          $self->{s_kwd} = '';
        }
        #
      } elsif ($self->{nc} == -1) {
        !!!cp (11);
        $self->{s_kwd} = '';
        !!!emit ({type => END_OF_FILE_TOKEN,
                  line => $self->{line}, column => $self->{column}});
        last A; ## TODO: ok?
      } else {
        !!!cp (12);
        $self->{s_kwd} = '';
        #
      }

      # Anything else
      my $token = {type => CHARACTER_TOKEN,
                   data => chr $self->{nc},
                   line => $self->{line}, column => $self->{column},
                  };
      if ($self->{read_until}->($token->{data}, q{-!<>&\]},
                                length $token->{data})) {
        $self->{s_kwd} = '';
      }

      ## Stay in the data state.
      if (not $self->{is_xml} and
          $self->{content_model} == PCDATA_CONTENT_MODEL) {
        !!!cp (13);
        $self->{state} = PCDATA_STATE;
      } else {
        !!!cp (14);
        ## Stay in the state.
      }
      !!!next-input-character;
      !!!emit ($token);
      redo A;
    } elsif ($self->{state} == TAG_OPEN_STATE) {
      ## XML5: "tag state".

      if ($self->{content_model} & CM_LIMITED_MARKUP) { # RCDATA | CDATA
        if ($self->{nc} == 0x002F) { # /
          !!!cp (15);
          !!!next-input-character;
          $self->{state} = CLOSE_TAG_OPEN_STATE;
          redo A;
        } elsif ($self->{nc} == 0x0021) { # !
          !!!cp (15.1);
          $self->{s_kwd} = $self->{escaped} ? '' : '<';
          #
        } else {
          !!!cp (16);
          $self->{s_kwd} = '';
          #
        }

        ## reconsume
        $self->{state} = DATA_STATE;
        !!!emit ({type => CHARACTER_TOKEN, data => '<',
                  line => $self->{line_prev},
                  column => $self->{column_prev},
                 });
        redo A;
      } elsif ($self->{content_model} & CM_FULL_MARKUP) { # PCDATA
        if ($self->{nc} == 0x0021) { # !
          !!!cp (17);
          $self->{state} = MARKUP_DECLARATION_OPEN_STATE;
          !!!next-input-character;
          redo A;
        } elsif ($self->{nc} == 0x002F) { # /
          !!!cp (18);
          $self->{state} = CLOSE_TAG_OPEN_STATE;
          !!!next-input-character;
          redo A;
        } elsif (0x0041 <= $self->{nc} and
                 $self->{nc} <= 0x005A) { # A..Z
          !!!cp (19);
          $self->{ct}
            = {type => START_TAG_TOKEN,
               tag_name => chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020)),
               line => $self->{line_prev},
               column => $self->{column_prev}};
          $self->{state} = TAG_NAME_STATE;
          !!!next-input-character;
          redo A;
        } elsif (0x0061 <= $self->{nc} and
                 $self->{nc} <= 0x007A) { # a..z
          !!!cp (20);
          $self->{ct} = {type => START_TAG_TOKEN,
                                    tag_name => chr ($self->{nc}),
                                    line => $self->{line_prev},
                                    column => $self->{column_prev}};
          $self->{state} = TAG_NAME_STATE;
          !!!next-input-character;
          redo A;
        } elsif ($self->{nc} == 0x003E) { # >
          !!!cp (21);
          !!!parse-error (type => 'empty start tag',
                          line => $self->{line_prev},
                          column => $self->{column_prev});
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          !!!next-input-character;

          !!!emit ({type => CHARACTER_TOKEN, data => '<>',
                    line => $self->{line_prev},
                    column => $self->{column_prev},
                   });

          redo A;
        } elsif ($self->{nc} == 0x003F) { # ?
          if ($self->{is_xml}) {
            !!!cp (22.1);
            $self->{state} = PI_STATE;
            !!!next-input-character;
            redo A;
          } else {
            !!!cp (22);
            !!!parse-error (type => 'pio',
                            line => $self->{line_prev},
                            column => $self->{column_prev});
            $self->{state} = BOGUS_COMMENT_STATE;
            $self->{ct} = {type => COMMENT_TOKEN, data => '',
                           line => $self->{line_prev},
                           column => $self->{column_prev},
                          };
            ## $self->{nc} is intentionally left as is
            redo A;
          }
        } elsif (not $self->{is_xml} or $is_space->{$self->{nc}}) {
          !!!cp (23);
          !!!parse-error (type => 'bare stago',
                          line => $self->{line_prev},
                          column => $self->{column_prev});
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume

          !!!emit ({type => CHARACTER_TOKEN, data => '<',
                    line => $self->{line_prev},
                    column => $self->{column_prev},
                   });

          redo A;
        } else {
          ## XML5: "<:" is a parse error.
          !!!cp (23.1);
          $self->{ct} = {type => START_TAG_TOKEN,
                                    tag_name => chr ($self->{nc}),
                                    line => $self->{line_prev},
                                    column => $self->{column_prev}};
          $self->{state} = TAG_NAME_STATE;
          !!!next-input-character;
          redo A;
        }
      } else {
        die "$0: $self->{content_model} in tag open";
      }
    } elsif ($self->{state} == CLOSE_TAG_OPEN_STATE) {
      ## NOTE: The "close tag open state" in the spec is implemented as
      ## |CLOSE_TAG_OPEN_STATE| and |CDATA_RCDATA_CLOSE_TAG_STATE|.

      ## XML5: "end tag state".

      my ($l, $c) = ($self->{line_prev}, $self->{column_prev} - 1); # "<"of"</"
      if ($self->{content_model} & CM_LIMITED_MARKUP) { # RCDATA | CDATA
        if (defined $self->{last_stag_name}) {
          $self->{state} = CDATA_RCDATA_CLOSE_TAG_STATE;
          $self->{kwd} = '';
          ## Reconsume.
          redo A;
        } else {
          ## No start tag token has ever been emitted
          ## NOTE: See <http://krijnhoetmer.nl/irc-logs/whatwg/20070626#l-564>.
          !!!cp (28);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## Reconsume.
          !!!emit ({type => CHARACTER_TOKEN, data => '</',
                    line => $l, column => $c,
                   });
          redo A;
        }
      }

      if (0x0041 <= $self->{nc} and
          $self->{nc} <= 0x005A) { # A..Z
        !!!cp (29);
        $self->{ct}
            = {type => END_TAG_TOKEN,
               tag_name => chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020)),
               line => $l, column => $c};
        $self->{state} = TAG_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif (0x0061 <= $self->{nc} and
               $self->{nc} <= 0x007A) { # a..z
        !!!cp (30);
        $self->{ct} = {type => END_TAG_TOKEN,
                                  tag_name => chr ($self->{nc}),
                                  line => $l, column => $c};
        $self->{state} = TAG_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'empty end tag',
                        line => $self->{line_prev}, ## "<" in "</>"
                        column => $self->{column_prev} - 1);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        if ($self->{is_xml}) {
          !!!cp (31);
          ## XML5: No parse error.
          
          ## NOTE: This parser raises a parse error, since it supports
          ## XML1, not XML5.

          ## NOTE: A short end tag token.
          my $ct = {type => END_TAG_TOKEN,
                    tag_name => '',
                    line => $self->{line_prev},
                    column => $self->{column_prev} - 1,
                   };
          !!!next-input-character;
          !!!emit ($ct);
        } else {
          !!!cp (31.1);
          !!!next-input-character;
        }
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (32);
        !!!parse-error (type => 'bare etago');
        $self->{s_kwd} = '';
        $self->{state} = DATA_STATE;
        # reconsume

        !!!emit ({type => CHARACTER_TOKEN, data => '</',
                  line => $l, column => $c,
                 });

        redo A;
      } elsif (not $self->{is_xml} or
               $is_space->{$self->{nc}}) {
        !!!cp (33);
        !!!parse-error (type => 'bogus end tag',
                        line => $self->{line_prev}, # "<" of "</"
                        column => $self->{column_prev} - 1);
        $self->{state} = BOGUS_COMMENT_STATE;
        $self->{ct} = {type => COMMENT_TOKEN, data => '',
                                  line => $self->{line_prev}, # "<" of "</"
                                  column => $self->{column_prev} - 1,
                                 };
        ## NOTE: $self->{nc} is intentionally left as is.
        ## Although the "anything else" case of the spec not explicitly
        ## states that the next input character is to be reconsumed,
        ## it will be included to the |data| of the comment token
        ## generated from the bogus end tag, as defined in the
        ## "bogus comment state" entry.
        redo A;
      } else {
        ## XML5: "</:" is a parse error.
        !!!cp (30.1);
        $self->{ct} = {type => END_TAG_TOKEN,
                       tag_name => chr ($self->{nc}),
                       line => $l, column => $c};
        $self->{state} = TAG_NAME_STATE; ## XML5: "end tag name state".
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == CDATA_RCDATA_CLOSE_TAG_STATE) {
      my $ch = substr $self->{last_stag_name}, length $self->{kwd}, 1;
      if (length $ch) {
        my $CH = $ch;
        $ch =~ tr/a-z/A-Z/;
        my $nch = chr $self->{nc};
        if ($nch eq $ch or $nch eq $CH) {
          !!!cp (24);
          ## Stay in the state.
          $self->{kwd} .= $nch;
          !!!next-input-character;
          redo A;
        } else {
          !!!cp (25);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## Reconsume.
          !!!emit ({type => CHARACTER_TOKEN,
                    data => '</' . $self->{kwd},
                    line => $self->{line_prev},
                    column => $self->{column_prev} - 1 - length $self->{kwd},
                   });
          redo A;
        }
      } else { # after "<{tag-name}"
        unless ($is_space->{$self->{nc}} or
                {
                 0x003E => 1, # >
                 0x002F => 1, # /
                 -1 => 1, # EOF
                }->{$self->{nc}}) {
          !!!cp (26);
          ## Reconsume.
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          !!!emit ({type => CHARACTER_TOKEN,
                    data => '</' . $self->{kwd},
                    line => $self->{line_prev},
                    column => $self->{column_prev} - 1 - length $self->{kwd},
                   });
          redo A;
        } else {
          !!!cp (27);
          $self->{ct}
              = {type => END_TAG_TOKEN,
                 tag_name => $self->{last_stag_name},
                 line => $self->{line_prev},
                 column => $self->{column_prev} - 1 - length $self->{kwd}};
          $self->{state} = TAG_NAME_STATE;
          ## Reconsume.
          redo A;
        }
      }
    } elsif ($self->{state} == TAG_NAME_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (34);
        $self->{state} = BEFORE_ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (35);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          #if ($self->{ct}->{attributes}) {
          #  ## NOTE: This should never be reached.
          #  !!! cp (36);
          #  !!! parse-error (type => 'end tag attribute');
          #} else {
            !!!cp (37);
          #}
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif (0x0041 <= $self->{nc} and
               $self->{nc} <= 0x005A) { # A..Z
        !!!cp (38);
        $self->{ct}->{tag_name}
            .= chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020));
          # start tag or end tag
        ## Stay in this state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (39);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          #if ($self->{ct}->{attributes}) {
          #  ## NOTE: This state should never be reached.
          #  !!! cp (40);
          #  !!! parse-error (type => 'end tag attribute');
          #} else {
            !!!cp (41);
          #}
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        # reconsume

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif ($self->{nc} == 0x002F) { # /
        !!!cp (42);
        $self->{state} = SELF_CLOSING_START_TAG_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (44);
        $self->{ct}->{tag_name} .= chr $self->{nc};
          # start tag or end tag
        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == BEFORE_ATTRIBUTE_NAME_STATE) {
      ## XML5: "Tag attribute name before state".

      if ($is_space->{$self->{nc}}) {
        !!!cp (45);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (46);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (47);
            !!!parse-error (type => 'end tag attribute');
          } else {
            !!!cp (48);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif (0x0041 <= $self->{nc} and
               $self->{nc} <= 0x005A) { # A..Z
        !!!cp (49);
        $self->{ca}
            = {name => chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020)),
               value => '',
               line => $self->{line}, column => $self->{column}};
        $self->{state} = ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x002F) { # /
        !!!cp (50);
        $self->{state} = SELF_CLOSING_START_TAG_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (52);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (53);
            !!!parse-error (type => 'end tag attribute');
          } else {
            !!!cp (54);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        # reconsume

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } else {
        if ({
             0x0022 => 1, # "
             0x0027 => 1, # '
             0x003D => 1, # =
            }->{$self->{nc}}) {
          !!!cp (55);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'bad attribute name');
        } else {
          !!!cp (56);
          ## XML5: ":" raises a parse error and is ignored.
        }
        $self->{ca}
            = {name => chr ($self->{nc}),
               value => '',
               line => $self->{line}, column => $self->{column}};
        $self->{state} = ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == ATTRIBUTE_NAME_STATE) {
      ## XML5: "Tag attribute name state".

      my $before_leave = sub {
        if (exists $self->{ct}->{attributes} # start tag or end tag
            ->{$self->{ca}->{name}}) { # MUST
          !!!cp (57);
          !!!parse-error (type => 'duplicate attribute', text => $self->{ca}->{name}, line => $self->{ca}->{line}, column => $self->{ca}->{column});
          ## Discard $self->{ca} # MUST
        } else {
          !!!cp (58);
          $self->{ct}->{attributes}->{$self->{ca}->{name}}
            = $self->{ca};
          $self->{ca}->{index} = ++$self->{ct}->{last_index};
        }
      }; # $before_leave

      if ($is_space->{$self->{nc}}) {
        !!!cp (59);
        $before_leave->();
        $self->{state} = AFTER_ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003D) { # =
        !!!cp (60);
        $before_leave->();
        $self->{state} = BEFORE_ATTRIBUTE_VALUE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{is_xml}) {
          !!!cp (60.1);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'no attr value'); ## TODO: type
        } else {
          !!!cp (60.2);
        }

        $before_leave->();
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (61);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          !!!cp (62);
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!parse-error (type => 'end tag attribute');
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif (0x0041 <= $self->{nc} and
               $self->{nc} <= 0x005A) { # A..Z
        !!!cp (63);
        $self->{ca}->{name}
            .= chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020));
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x002F) { # /
        if ($self->{is_xml}) {
          !!!cp (64);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'no attr value'); ## TODO: type
        } else {
          !!!cp (64.1);
        }
        
        $before_leave->();
        $self->{state} = SELF_CLOSING_START_TAG_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        $before_leave->();
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (66);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (67);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (68);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        # reconsume

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } else {
        if ($self->{nc} == 0x0022 or # "
            $self->{nc} == 0x0027) { # '
          !!!cp (69);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'bad attribute name');
        } else {
          !!!cp (70);
        }
        $self->{ca}->{name} .= chr ($self->{nc});
        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_ATTRIBUTE_NAME_STATE) {
      ## XML5: "Tag attribute name after state".
      
      if ($is_space->{$self->{nc}}) {
        !!!cp (71);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003D) { # =
        !!!cp (72);
        $self->{state} = BEFORE_ATTRIBUTE_VALUE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{is_xml}) {
          !!!cp (72.1);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'no attr value'); ## TODO: type
        } else {
          !!!cp (72.2);
        }

        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (73);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (74);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (75);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif (0x0041 <= $self->{nc} and
               $self->{nc} <= 0x005A) { # A..Z
        !!!cp (76);
        $self->{ca}
            = {name => chr ($self->{nc} + ($self->{is_xml} ? 0 : 0x0020)),
               value => '',
               line => $self->{line}, column => $self->{column}};
        $self->{state} = ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x002F) { # /
        if ($self->{is_xml}) {
          !!!cp (77);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'no attr value'); ## TODO: type
        } else {
          !!!cp (77.1);
        }
        
        $self->{state} = SELF_CLOSING_START_TAG_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (79);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (80);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (81);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{s_kwd} = '';
        $self->{state} = DATA_STATE;
        # reconsume

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } else {
        if ($self->{is_xml}) {
          !!!cp (78.1);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'no attr value'); ## TODO: type
        } else {
          !!!cp (78.2);
        }

        if ($self->{nc} == 0x0022 or # "
            $self->{nc} == 0x0027) { # '
          !!!cp (78);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'bad attribute name');
        } else {
          !!!cp (82);
        }
        $self->{ca}
            = {name => chr ($self->{nc}),
               value => '',
               line => $self->{line}, column => $self->{column}};
        $self->{state} = ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;        
      }
    } elsif ($self->{state} == BEFORE_ATTRIBUTE_VALUE_STATE) {
      ## XML5: "Tag attribute value before state".

      if ($is_space->{$self->{nc}}) {
        !!!cp (83);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        !!!cp (84);
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0026) { # &
        !!!cp (85);
        $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        ## reconsume
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        !!!cp (86);
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'empty unquoted attribute value');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (87);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (88);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (89);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (90);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (91);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (92);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } else {
        if ($self->{nc} == 0x003D) { # =
          !!!cp (93);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'bad attribute value');
        } elsif ($self->{is_xml}) {
          !!!cp (93.1);
          ## XML5: No parse error.
          !!!parse-error (type => 'unquoted attr value'); ## TODO
        } else {
          !!!cp (94);
        }
        $self->{ca}->{value} .= chr ($self->{nc});
        $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE) {
      ## XML5: "Tag attribute value double quoted state" and "DOCTYPE
      ## ATTLIST attribute value double quoted state".
      
      if ($self->{nc} == 0x0022) { # "
        if ($self->{ct}->{type} == ATTLIST_TOKEN) {
          !!!cp (95.1);
          ## XML5: "DOCTYPE ATTLIST name after state".
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = AFTER_ATTLIST_ATTR_VALUE_QUOTED_STATE;
        } else {
          !!!cp (95);
          ## XML5: "Tag attribute name before state".
          $self->{state} = AFTER_ATTRIBUTE_VALUE_QUOTED_STATE;
        }
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0026) { # &
        !!!cp (96);
        ## XML5: Not defined yet.

        ## NOTE: In the spec, the tokenizer is switched to the 
        ## "entity in attribute value state".  In this implementation, the
        ## tokenizer is switched to the |ENTITY_STATE|, which is an
        ## implementation of the "consume a character reference" algorithm.
        $self->{prev_state} = $self->{state};
        $self->{entity_add} = 0x0022; # "
        $self->{state} = ENTITY_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed attribute value');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (97);
          $self->{last_stag_name} = $self->{ct}->{tag_name};

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # start tag
          redo A;
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (98);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (99);
          }

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # end tag
          redo A;
        } elsif ($self->{ct}->{type} == ATTLIST_TOKEN) {
          ## XML5: No parse error above; not defined yet.
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
          ## Reconsume.
          !!!emit ($self->{ct}); # ATTLIST
          redo A;
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
      } else {
        ## XML5 [ATTLIST]: Not defined yet.
        if ($self->{is_xml} and $self->{nc} == 0x003C) { # <
          !!!cp (100);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'lt in attr value'); ## TODO: type
        } else {
          !!!cp (100.1);
        }
        $self->{ca}->{value} .= chr ($self->{nc});
        $self->{read_until}->($self->{ca}->{value},
                              q["&<],
                              length $self->{ca}->{value});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE) {
      ## XML5: "Tag attribute value single quoted state" and "DOCTYPE
      ## ATTLIST attribute value single quoted state".

      if ($self->{nc} == 0x0027) { # '
        if ($self->{ct}->{type} == ATTLIST_TOKEN) {
          !!!cp (101.1);
          ## XML5: "DOCTYPE ATTLIST name after state".
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = AFTER_ATTLIST_ATTR_VALUE_QUOTED_STATE;
        } else {
          !!!cp (101);
          ## XML5: "Before attribute name state" (sic).
          $self->{state} = AFTER_ATTRIBUTE_VALUE_QUOTED_STATE;
        }
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0026) { # &
        !!!cp (102);
        ## XML5: Not defined yet.

        ## NOTE: In the spec, the tokenizer is switched to the 
        ## "entity in attribute value state".  In this implementation, the
        ## tokenizer is switched to the |ENTITY_STATE|, which is an
        ## implementation of the "consume a character reference" algorithm.
        $self->{entity_add} = 0x0027; # '
        $self->{prev_state} = $self->{state};
        $self->{state} = ENTITY_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed attribute value');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (103);
          $self->{last_stag_name} = $self->{ct}->{tag_name};

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # start tag
          redo A;
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (104);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (105);
          }

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # end tag
          redo A;
        } elsif ($self->{ct}->{type} == ATTLIST_TOKEN) {
          ## XML5: No parse error above; not defined yet.
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
          ## Reconsume.
          !!!emit ($self->{ct}); # ATTLIST
          redo A;
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
      } else {
        ## XML5 [ATTLIST]: Not defined yet.
        if ($self->{is_xml} and $self->{nc} == 0x003C) { # <
          !!!cp (106);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'lt in attr value'); ## TODO: type
        } else {
          !!!cp (106.1);
        }
        $self->{ca}->{value} .= chr ($self->{nc});
        $self->{read_until}->($self->{ca}->{value},
                              q['&<],
                              length $self->{ca}->{value});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == ATTRIBUTE_VALUE_UNQUOTED_STATE) {
      ## XML5: "Tag attribute value unquoted state".

      if ($is_space->{$self->{nc}}) {
        if ($self->{ct}->{type} == ATTLIST_TOKEN) {
          !!!cp (107.1);
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_ATTLIST_NAME_AFTER_STATE;
        } else {
          !!!cp (107);
          ## XML5: "Tag attribute name before state".
          $self->{state} = BEFORE_ATTRIBUTE_NAME_STATE;
        }
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0026) { # &
        !!!cp (108);

        ## XML5: Not defined yet.

        ## NOTE: In the spec, the tokenizer is switched to the 
        ## "entity in attribute value state".  In this implementation, the
        ## tokenizer is switched to the |ENTITY_STATE|, which is an
        ## implementation of the "consume a character reference" algorithm.
        $self->{entity_add} = -1;
        $self->{prev_state} = $self->{state};
        $self->{state} = ENTITY_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (109);
          $self->{last_stag_name} = $self->{ct}->{tag_name};

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          !!!next-input-character;
          !!!emit ($self->{ct}); # start tag
          redo A;
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (110);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (111);
          }

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          !!!next-input-character;
          !!!emit ($self->{ct}); # end tag
          redo A;
        } elsif ($self->{ct}->{type} == ATTLIST_TOKEN) {
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
          !!!next-input-character;
          !!!emit ($self->{ct}); # ATTLIST
          redo A;
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
      } elsif ($self->{nc} == -1) {
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (112);
          !!!parse-error (type => 'unclosed tag');
          $self->{last_stag_name} = $self->{ct}->{tag_name};

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # start tag
          redo A;
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          !!!parse-error (type => 'unclosed tag');
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (113);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (114);
          }

          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
          ## reconsume
          !!!emit ($self->{ct}); # end tag
          redo A;
        } elsif ($self->{ct}->{type} == ATTLIST_TOKEN) {
          !!!parse-error (type => 'unclosed md'); ## TODO: type
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
          ## Reconsume.
          !!!emit ($self->{ct}); # ATTLIST
          redo A;
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
      } else {
        if ({
             0x0022 => 1, # "
             0x0027 => 1, # '
             0x003D => 1, # =
            }->{$self->{nc}}) {
          !!!cp (115);
          ## XML5: Not a parse error.
          !!!parse-error (type => 'bad attribute value');
        } else {
          !!!cp (116);
        }
        $self->{ca}->{value} .= chr ($self->{nc});
        $self->{read_until}->($self->{ca}->{value},
                              q["'=& >],
                              length $self->{ca}->{value});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_ATTRIBUTE_VALUE_QUOTED_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (118);
        $self->{state} = BEFORE_ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (119);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp (120);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (121);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif ($self->{nc} == 0x002F) { # /
        !!!cp (122);
        $self->{state} = SELF_CLOSING_START_TAG_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag'); 
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (122.3);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          if ($self->{ct}->{attributes}) {
            !!!cp (122.1);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (122.2);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ($self->{ct}); # start tag or end tag
        redo A;
      } else {
        !!!cp ('124.1');
        !!!parse-error (type => 'no space between attributes');
        $self->{state} = BEFORE_ATTRIBUTE_NAME_STATE;
        ## reconsume
        redo A;
      }
    } elsif ($self->{state} == SELF_CLOSING_START_TAG_STATE) {
      ## XML5: "Empty tag state".

      if ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == END_TAG_TOKEN) {
          !!!cp ('124.2');
          !!!parse-error (type => 'nestc', token => $self->{ct});
          ## TODO: Different type than slash in start tag
          $self->{content_model} = PCDATA_CONTENT_MODEL; # MUST
          if ($self->{ct}->{attributes}) {
            !!!cp ('124.4');
            !!!parse-error (type => 'end tag attribute');
          } else {
            !!!cp ('124.5');
          }
          ## TODO: Test |<title></title/>|
        } else {
          !!!cp ('124.3');
          $self->{self_closing} = 1;
        }

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # start tag or end tag

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed tag');
        if ($self->{ct}->{type} == START_TAG_TOKEN) {
          !!!cp (124.7);
          $self->{last_stag_name} = $self->{ct}->{tag_name};
        } elsif ($self->{ct}->{type} == END_TAG_TOKEN) {
          if ($self->{ct}->{attributes}) {
            !!!cp (124.5);
            !!!parse-error (type => 'end tag attribute');
          } else {
            ## NOTE: This state should never be reached.
            !!!cp (124.6);
          }
        } else {
          die "$0: $self->{ct}->{type}: Unknown token type";
        }
        ## XML5: "Tag attribute name before state".
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ($self->{ct}); # start tag or end tag
        redo A;
      } else {
        !!!cp ('124.4');
        !!!parse-error (type => 'nestc');
        ## TODO: This error type is wrong.
        $self->{state} = BEFORE_ATTRIBUTE_NAME_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == BOGUS_COMMENT_STATE) {
      ## XML5: "Bogus comment state" and "DOCTYPE bogus comment state".

      ## NOTE: Unlike spec's "bogus comment state", this implementation
      ## consumes characters one-by-one basis.
      
      if ($self->{nc} == 0x003E) { # >
        if ($self->{in_subset}) {
          !!!cp (123);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (124);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;

        !!!emit ($self->{ct}); # comment
        redo A;
      } elsif ($self->{nc} == -1) { 
        if ($self->{in_subset}) {
          !!!cp (125.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (125);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment
        redo A;
      } else {
        !!!cp (126);
        $self->{ct}->{data} .= chr ($self->{nc}); # comment
        $self->{read_until}->($self->{ct}->{data},
                              q[>],
                              length $self->{ct}->{data});

        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == MARKUP_DECLARATION_OPEN_STATE) {
      ## XML5: "Markup declaration state".
      
      if ($self->{nc} == 0x002D) { # -
        !!!cp (133);
        $self->{state} = MD_HYPHEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0044 or # D
               $self->{nc} == 0x0064) { # d
        ## ASCII case-insensitive.
        !!!cp (130);
        $self->{state} = MD_DOCTYPE_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ((($self->{insertion_mode} & IN_FOREIGN_CONTENT_IM and
                 $self->{open_elements}->[-1]->[1] & FOREIGN_EL) or
                $self->{is_xml}) and
               $self->{nc} == 0x005B) { # [
        !!!cp (135.4);                
        $self->{state} = MD_CDATA_STATE;
        $self->{kwd} = '[';
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (136);
      }

      !!!parse-error (type => 'bogus comment',
                      line => $self->{line_prev},
                      column => $self->{column_prev} - 1);
      ## Reconsume.
      $self->{state} = BOGUS_COMMENT_STATE;
      $self->{ct} = {type => COMMENT_TOKEN, data => '',
                                line => $self->{line_prev},
                                column => $self->{column_prev} - 1,
                               };
      redo A;
    } elsif ($self->{state} == MD_HYPHEN_STATE) {
      if ($self->{nc} == 0x002D) { # -
        !!!cp (127);
        $self->{ct} = {type => COMMENT_TOKEN, data => '',
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 2,
                                 };
        $self->{state} = COMMENT_START_STATE; ## XML5: "comment state".
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (128);
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 2);
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN,
                                  data => '-',
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 2,
                                 };
        redo A;
      }
    } elsif ($self->{state} == MD_DOCTYPE_STATE) {
      ## ASCII case-insensitive.
      if ($self->{nc} == [
            undef,
            0x004F, # O
            0x0043, # C
            0x0054, # T
            0x0059, # Y
            0x0050, # P
          ]->[length $self->{kwd}] or
          $self->{nc} == [
            undef,
            0x006F, # o
            0x0063, # c
            0x0074, # t
            0x0079, # y
            0x0070, # p
          ]->[length $self->{kwd}]) {
        !!!cp (131);
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ((length $self->{kwd}) == 6 and
               ($self->{nc} == 0x0045 or # E
                $self->{nc} == 0x0065)) { # e
        if ($self->{is_xml} and
            ($self->{kwd} ne 'DOCTYP' or $self->{nc} == 0x0065)) {
          !!!cp (129);
          ## XML5: case-sensitive.
          !!!parse-error (type => 'lowercase keyword', ## TODO
                          text => 'DOCTYPE',
                          line => $self->{line_prev},
                          column => $self->{column_prev} - 5);
        } else {
          !!!cp (129.1);
        }
        $self->{state} = DOCTYPE_STATE;
        $self->{ct} = {type => DOCTYPE_TOKEN,
                                  quirks => 1,
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 7,
                                 };
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (132);        
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1 - length $self->{kwd});
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN,
                                  data => $self->{kwd},
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 1 - length $self->{kwd},
                                 };
        redo A;
      }
    } elsif ($self->{state} == MD_CDATA_STATE) {
      if ($self->{nc} == {
            '[' => 0x0043, # C
            '[C' => 0x0044, # D
            '[CD' => 0x0041, # A
            '[CDA' => 0x0054, # T
            '[CDAT' => 0x0041, # A
          }->{$self->{kwd}}) {
        !!!cp (135.1);
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{kwd} eq '[CDATA' and
               $self->{nc} == 0x005B) { # [
        if ($self->{is_xml} and 
            not $self->{tainted} and
            @{$self->{open_elements} or []} == 0) {
          !!!cp (135.2);
          !!!parse-error (type => 'cdata outside of root element',
                          line => $self->{line_prev},
                          column => $self->{column_prev} - 7);
          $self->{tainted} = 1;
        } else {
          !!!cp (135.21);
        }

        $self->{ct} = {type => CHARACTER_TOKEN,
                                  data => '',
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 7};
        $self->{state} = CDATA_SECTION_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (135.3);
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1 - length $self->{kwd});
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN,
                                  data => $self->{kwd},
                                  line => $self->{line_prev},
                                  column => $self->{column_prev} - 1 - length $self->{kwd},
                                 };
        redo A;
      }
    } elsif ($self->{state} == COMMENT_START_STATE) {
      if ($self->{nc} == 0x002D) { # -
        !!!cp (137);
        $self->{state} = COMMENT_START_DASH_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'bogus comment');
        if ($self->{in_subset}) {
          !!!cp (138.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (138);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;

        !!!emit ($self->{ct}); # comment

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed comment');
        if ($self->{in_subset}) {
          !!!cp (139.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (139);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment

        redo A;
      } else {
        !!!cp (140);
        $self->{ct}->{data} # comment
            .= chr ($self->{nc});
        $self->{state} = COMMENT_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == COMMENT_START_DASH_STATE) {
      if ($self->{nc} == 0x002D) { # -
        !!!cp (141);
        $self->{state} = COMMENT_END_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'bogus comment');
        if ($self->{in_subset}) {
          !!!cp (142.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (142);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;

        !!!emit ($self->{ct}); # comment

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed comment');
        if ($self->{in_subset}) {
          !!!cp (143.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (143);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment

        redo A;
      } else {
        !!!cp (144);
        $self->{ct}->{data} # comment
            .= '-' . chr ($self->{nc});
        $self->{state} = COMMENT_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == COMMENT_STATE) {
      ## XML5: "Comment state" and "DOCTYPE comment state".

      if ($self->{nc} == 0x002D) { # -
        !!!cp (145);
        $self->{state} = COMMENT_END_DASH_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed comment');
        if ($self->{in_subset}) {
          !!!cp (146.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (146);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment

        redo A;
      } else {
        !!!cp (147);
        $self->{ct}->{data} .= chr ($self->{nc}); # comment
        $self->{read_until}->($self->{ct}->{data},
                              q[-],
                              length $self->{ct}->{data});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == COMMENT_END_DASH_STATE) {
      ## XML5: "Comment dash state" and "DOCTYPE comment dash state".

      if ($self->{nc} == 0x002D) { # -
        !!!cp (148);
        $self->{state} = COMMENT_END_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed comment');
        if ($self->{in_subset}) {
          !!!cp (149.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (149);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment

        redo A;
      } else {
        !!!cp (150);
        $self->{ct}->{data} .= '-' . chr ($self->{nc}); # comment
        $self->{state} = COMMENT_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == COMMENT_END_STATE) {
      ## XML5: "Comment end state" and "DOCTYPE comment end state".

      if ($self->{nc} == 0x003E) { # >
        if ($self->{in_subset}) {
          !!!cp (151.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (151);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;

        !!!emit ($self->{ct}); # comment

        redo A;
      } elsif ($self->{nc} == 0x002D) { # -
        !!!cp (152);
        ## XML5: Not a parse error.
        !!!parse-error (type => 'dash in comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev});
        $self->{ct}->{data} .= '-'; # comment
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed comment');
        if ($self->{in_subset}) {
          !!!cp (153.1);
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          !!!cp (153);
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## reconsume

        !!!emit ($self->{ct}); # comment

        redo A;
      } else {
        !!!cp (154);
        ## XML5: Not a parse error.
        !!!parse-error (type => 'dash in comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev});
        $self->{ct}->{data} .= '--' . chr ($self->{nc}); # comment
        $self->{state} = COMMENT_STATE;
        !!!next-input-character;
        redo A;
      } 
    } elsif ($self->{state} == DOCTYPE_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (155);
        $self->{state} = BEFORE_DOCTYPE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (156);
        ## XML5: Unless EOF, swith to the bogus comment state.
        !!!parse-error (type => 'no space before DOCTYPE name');
        $self->{state} = BEFORE_DOCTYPE_NAME_STATE;
        ## reconsume
        redo A;
      }
    } elsif ($self->{state} == BEFORE_DOCTYPE_NAME_STATE) {
      ## XML5: "DOCTYPE root name before state".

      if ($is_space->{$self->{nc}}) {
        !!!cp (157);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (158);
        ## XML5: No parse error.
        !!!parse-error (type => 'no DOCTYPE name');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE (quirks)

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (159);
        !!!parse-error (type => 'no DOCTYPE name');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        !!!emit ($self->{ct}); # DOCTYPE (quirks)

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (159.1);
        !!!parse-error (type => 'no DOCTYPE name');
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (160);
        $self->{ct}->{name} = chr $self->{nc};
        delete $self->{ct}->{quirks};
        $self->{state} = DOCTYPE_NAME_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_NAME_STATE) {
      ## XML5: "DOCTYPE root name state".

      ## ISSUE: Redundant "First," in the spec.

      if ($is_space->{$self->{nc}}) {
        !!!cp (161);
        $self->{state} = AFTER_DOCTYPE_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (162);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (163);
        !!!parse-error (type => 'unclosed DOCTYPE');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (163.1);
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (164);
        $self->{ct}->{name}
          .= chr ($self->{nc}); # DOCTYPE
        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_DOCTYPE_NAME_STATE) {
      ## XML5: Corresponding to XML5's "DOCTYPE root name after
      ## state", but implemented differently.

      if ($is_space->{$self->{nc}}) {
        !!!cp (165);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (166);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (167);
        !!!parse-error (type => 'unclosed DOCTYPE');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == 0x0050 or # P
               $self->{nc} == 0x0070) { # p
        !!!cp (167.1);
        $self->{state} = PUBLIC_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0053 or # S
               $self->{nc} == 0x0073) { # s
        !!!cp (167.2);
        $self->{state} = SYSTEM_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (167.3);
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (180);
        !!!parse-error (type => 'string after DOCTYPE name');
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == PUBLIC_STATE) {
      ## ASCII case-insensitive
      if ($self->{nc} == [
            undef, 
            0x0055, # U
            0x0042, # B
            0x004C, # L
            0x0049, # I
          ]->[length $self->{kwd}] or
          $self->{nc} == [
            undef, 
            0x0075, # u
            0x0062, # b
            0x006C, # l
            0x0069, # i
          ]->[length $self->{kwd}]) {
        !!!cp (175);
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ((length $self->{kwd}) == 5 and
               ($self->{nc} == 0x0043 or # C
                $self->{nc} == 0x0063)) { # c
        if ($self->{is_xml} and
            ($self->{kwd} ne 'PUBLI' or $self->{nc} == 0x0063)) { # c
          !!!cp (168.1);
          !!!parse-error (type => 'lowercase keyword', ## TODO: type
                          text => 'PUBLIC',
                          line => $self->{line_prev},
                          column => $self->{column_prev} - 4);
        } else {
          !!!cp (168);
        }
        $self->{state} = BEFORE_DOCTYPE_PUBLIC_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (169);
        !!!parse-error (type => 'string after DOCTYPE name',
                        line => $self->{line_prev},
                        column => $self->{column_prev} + 1 - length $self->{kwd});
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == SYSTEM_STATE) {
      ## ASCII case-insensitive
      if ($self->{nc} == [
            undef, 
            0x0059, # Y
            0x0053, # S
            0x0054, # T
            0x0045, # E
          ]->[length $self->{kwd}] or
          $self->{nc} == [
            undef, 
            0x0079, # y
            0x0073, # s
            0x0074, # t
            0x0065, # e
          ]->[length $self->{kwd}]) {
        !!!cp (170);
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ((length $self->{kwd}) == 5 and
               ($self->{nc} == 0x004D or # M
                $self->{nc} == 0x006D)) { # m
        if ($self->{is_xml} and
            ($self->{kwd} ne 'SYSTE' or $self->{nc} == 0x006D)) { # m
          !!!cp (171.1);
          !!!parse-error (type => 'lowercase keyword', ## TODO: type
                          text => 'SYSTEM',
                          line => $self->{line_prev},
                          column => $self->{column_prev} - 4);
        } else {
          !!!cp (171);
        }
        $self->{state} = BEFORE_DOCTYPE_SYSTEM_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (172);
        !!!parse-error (type => 'string after DOCTYPE name',
                        line => $self->{line_prev},
                        column => $self->{column_prev} + 1 - length $self->{kwd});
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == BEFORE_DOCTYPE_PUBLIC_IDENTIFIER_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (181);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} eq 0x0022) { # "
        !!!cp (182);
        $self->{ct}->{pubid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_PUBLIC_IDENTIFIER_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} eq 0x0027) { # '
        !!!cp (183);
        $self->{ct}->{pubid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_PUBLIC_IDENTIFIER_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} eq 0x003E) { # >
        !!!cp (184);
        !!!parse-error (type => 'no PUBLIC literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (185);
        !!!parse-error (type => 'unclosed DOCTYPE');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (186.1);
        !!!parse-error (type => 'no PUBLIC literal');
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (186);
        !!!parse-error (type => 'string after PUBLIC');
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_PUBLIC_IDENTIFIER_DOUBLE_QUOTED_STATE) {
      if ($self->{nc} == 0x0022) { # "
        !!!cp (187);
        $self->{state} = AFTER_DOCTYPE_PUBLIC_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (188);
        !!!parse-error (type => 'unclosed PUBLIC literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (189);
        !!!parse-error (type => 'unclosed PUBLIC literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } else {
        !!!cp (190);
        $self->{ct}->{pubid} # DOCTYPE
            .= chr $self->{nc};
        $self->{read_until}->($self->{ct}->{pubid}, q[">],
                              length $self->{ct}->{pubid});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_PUBLIC_IDENTIFIER_SINGLE_QUOTED_STATE) {
      if ($self->{nc} == 0x0027) { # '
        !!!cp (191);
        $self->{state} = AFTER_DOCTYPE_PUBLIC_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (192);
        !!!parse-error (type => 'unclosed PUBLIC literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (193);
        !!!parse-error (type => 'unclosed PUBLIC literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } else {
        !!!cp (194);
        $self->{ct}->{pubid} # DOCTYPE
            .= chr $self->{nc};
        $self->{read_until}->($self->{ct}->{pubid}, q['>],
                              length $self->{ct}->{pubid});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_DOCTYPE_PUBLIC_IDENTIFIER_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (195);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        !!!cp (196);
        $self->{ct}->{sysid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_SYSTEM_IDENTIFIER_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        !!!cp (197);
        $self->{ct}->{sysid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_SYSTEM_IDENTIFIER_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{is_xml}) {
          !!!cp (198.1);
          !!!parse-error (type => 'no SYSTEM literal');
        } else {
          !!!cp (198);
        }
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (199);
        !!!parse-error (type => 'unclosed DOCTYPE');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (200.1);
        !!!parse-error (type => 'no SYSTEM literal');
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (200);
        !!!parse-error (type => 'string after PUBLIC literal');
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == BEFORE_DOCTYPE_SYSTEM_IDENTIFIER_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (201);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        !!!cp (202);
        $self->{ct}->{sysid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_SYSTEM_IDENTIFIER_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        !!!cp (203);
        $self->{ct}->{sysid} = ''; # DOCTYPE
        $self->{state} = DOCTYPE_SYSTEM_IDENTIFIER_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (204);
        !!!parse-error (type => 'no SYSTEM literal');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (205);
        !!!parse-error (type => 'unclosed DOCTYPE');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (206.1);
        !!!parse-error (type => 'no SYSTEM literal');

        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (206);
        !!!parse-error (type => 'string after SYSTEM');
        $self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_SYSTEM_IDENTIFIER_DOUBLE_QUOTED_STATE) {
      if ($self->{nc} == 0x0022) { # "
        !!!cp (207);
        $self->{state} = AFTER_DOCTYPE_SYSTEM_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } elsif (not $self->{is_xml} and $self->{nc} == 0x003E) { # >
        !!!cp (208);
        !!!parse-error (type => 'unclosed SYSTEM literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (209);
        !!!parse-error (type => 'unclosed SYSTEM literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } else {
        !!!cp (210);
        $self->{ct}->{sysid} # DOCTYPE
            .= chr $self->{nc};
        $self->{read_until}->($self->{ct}->{sysid}, q[">],
                              length $self->{ct}->{sysid});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_SYSTEM_IDENTIFIER_SINGLE_QUOTED_STATE) {
      if ($self->{nc} == 0x0027) { # '
        !!!cp (211);
        $self->{state} = AFTER_DOCTYPE_SYSTEM_IDENTIFIER_STATE;
        !!!next-input-character;
        redo A;
      } elsif (not $self->{is_xml} and $self->{nc} == 0x003E) { # >
        !!!cp (212);
        !!!parse-error (type => 'unclosed SYSTEM literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (213);
        !!!parse-error (type => 'unclosed SYSTEM literal');

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } else {
        !!!cp (214);
        $self->{ct}->{sysid} # DOCTYPE
            .= chr $self->{nc};
        $self->{read_until}->($self->{ct}->{sysid}, q['>],
                              length $self->{ct}->{sysid});

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_DOCTYPE_SYSTEM_IDENTIFIER_STATE) {
      if ($is_space->{$self->{nc}}) {
        !!!cp (215);
        ## Stay in the state
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!cp (216);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (217);
        !!!parse-error (type => 'unclosed DOCTYPE');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        $self->{ct}->{quirks} = 1;
        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (218.1);
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } else {
        !!!cp (218);
        !!!parse-error (type => 'string after SYSTEM literal');
        #$self->{ct}->{quirks} = 1;

        $self->{state} = BOGUS_DOCTYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == BOGUS_DOCTYPE_STATE) {
      if ($self->{nc} == 0x003E) { # >
        !!!cp (219);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } elsif ($self->{is_xml} and $self->{nc} == 0x005B) { # [
        !!!cp (220.1);
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        $self->{ct}->{has_internal_subset} = 1; # DOCTYPE
        $self->{in_subset} = 1;
        !!!next-input-character;
        !!!emit ($self->{ct}); # DOCTYPE
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!cp (220);
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## reconsume

        !!!emit ($self->{ct}); # DOCTYPE

        redo A;
      } else {
        !!!cp (221);
        my $s = '';
        $self->{read_until}->($s, q{>[}, 0);

        ## Stay in the state
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == CDATA_SECTION_STATE) {
      ## NOTE: "CDATA section state" in the state is jointly implemented
      ## by three states, |CDATA_SECTION_STATE|, |CDATA_SECTION_MSE1_STATE|,
      ## and |CDATA_SECTION_MSE2_STATE|.

      ## XML5: "CDATA state".
      
      if ($self->{nc} == 0x005D) { # ]
        !!!cp (221.1);
        $self->{state} = CDATA_SECTION_MSE1_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        if ($self->{is_xml}) {
          !!!cp (221.11);
          !!!parse-error (type => 'no mse'); ## TODO: type
        } else {
          !!!cp (221.12);
        }

        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        if (length $self->{ct}->{data}) { # character
          !!!cp (221.2);
          !!!emit ($self->{ct}); # character
        } else {
          !!!cp (221.3);
          ## No token to emit. $self->{ct} is discarded.
        }        
        redo A;
      } else {
        !!!cp (221.4);
        $self->{ct}->{data} .= chr $self->{nc};
        $self->{read_until}->($self->{ct}->{data},
                              q<]>,
                              length $self->{ct}->{data});

        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }

      ## ISSUE: "text tokens" in spec.
    } elsif ($self->{state} == CDATA_SECTION_MSE1_STATE) {
      ## XML5: "CDATA bracket state".

      if ($self->{nc} == 0x005D) { # ]
        !!!cp (221.5);
        $self->{state} = CDATA_SECTION_MSE2_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (221.6);
        ## XML5: If EOF, "]" is not appended and changed to the data state.
        $self->{ct}->{data} .= ']';
        $self->{state} = CDATA_SECTION_STATE; ## XML5: Stay in the state.
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == CDATA_SECTION_MSE2_STATE) {
      ## XML5: "CDATA end state".

      if ($self->{nc} == 0x003E) { # >
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;
        if (length $self->{ct}->{data}) { # character
          !!!cp (221.7);
          !!!emit ($self->{ct}); # character
        } else {
          !!!cp (221.8);
          ## No token to emit. $self->{ct} is discarded.
        }
        redo A;
      } elsif ($self->{nc} == 0x005D) { # ]
        !!!cp (221.9); # character
        $self->{ct}->{data} .= ']'; ## Add first "]" of "]]]".
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (221.11);
        $self->{ct}->{data} .= ']]'; # character
        $self->{state} = CDATA_SECTION_STATE;
        ## Reconsume. ## XML5: Emit.
        redo A;
      }
    } elsif ($self->{state} == ENTITY_STATE) {
      if ($is_space->{$self->{nc}} or
          {
            0x003C => 1, 0x0026 => 1, -1 => 1, # <, &
            $self->{entity_add} => 1,
          }->{$self->{nc}}) {
        !!!cp (1001);
        ## Don't consume
        ## No error
        ## Return nothing.
        #
      } elsif ($self->{nc} == 0x0023) { # #
        !!!cp (999);
        $self->{state} = ENTITY_HASH_STATE;
        $self->{kwd} = '#';
        !!!next-input-character;
        redo A;
      } elsif ((0x0041 <= $self->{nc} and
                $self->{nc} <= 0x005A) or # A..Z
               (0x0061 <= $self->{nc} and
                $self->{nc} <= 0x007A)) { # a..z
        !!!cp (998);
        require Whatpm::_NamedEntityList;
        $self->{state} = ENTITY_NAME_STATE;
        $self->{kwd} = chr $self->{nc};
        $self->{entity__value} = $self->{kwd};
        $self->{entity__match} = 0;
        !!!next-input-character;
        redo A;
      } else {
        !!!cp (1027);
        !!!parse-error (type => 'bare ero');
        ## Return nothing.
        #
      }

      ## NOTE: No character is consumed by the "consume a character
      ## reference" algorithm.  In other word, there is an "&" character
      ## that does not introduce a character reference, which would be
      ## appended to the parent element or the attribute value in later
      ## process of the tokenizer.

      if ($self->{prev_state} == DATA_STATE) {
        !!!cp (997);
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => CHARACTER_TOKEN, data => '&',
                  line => $self->{line_prev},
                  column => $self->{column_prev},
                 });
        redo A;
      } else {
        !!!cp (996);
        $self->{ca}->{value} .= '&';
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == ENTITY_HASH_STATE) {
      if ($self->{nc} == 0x0078 or # x
          $self->{nc} == 0x0058) { # X
        !!!cp (995);
        $self->{state} = HEXREF_X_STATE;
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif (0x0030 <= $self->{nc} and
               $self->{nc} <= 0x0039) { # 0..9
        !!!cp (994);
        $self->{state} = NCR_NUM_STATE;
        $self->{kwd} = $self->{nc} - 0x0030;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'bare nero',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1);

        ## NOTE: According to the spec algorithm, nothing is returned,
        ## and then "&#" is appended to the parent element or the attribute 
        ## value in the later processing.

        if ($self->{prev_state} == DATA_STATE) {
          !!!cp (1019);
          $self->{state} = $self->{prev_state};
          $self->{s_kwd} = '';
          ## Reconsume.
          !!!emit ({type => CHARACTER_TOKEN,
                    data => '&#',
                    line => $self->{line_prev},
                    column => $self->{column_prev} - 1,
                   });
          redo A;
        } else {
          !!!cp (993);
          $self->{ca}->{value} .= '&#';
          $self->{state} = $self->{prev_state};
          $self->{s_kwd} = '';
          ## Reconsume.
          redo A;
        }
      }
    } elsif ($self->{state} == NCR_NUM_STATE) {
      if (0x0030 <= $self->{nc} and 
          $self->{nc} <= 0x0039) { # 0..9
        !!!cp (1012);
        $self->{kwd} *= 10;
        $self->{kwd} += $self->{nc} - 0x0030;
        
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003B) { # ;
        !!!cp (1013);
        !!!next-input-character;
        #
      } else {
        !!!cp (1014);
        !!!parse-error (type => 'no refc');
        ## Reconsume.
        #
      }

      my $code = $self->{kwd};
      my $l = $self->{line_prev};
      my $c = $self->{column_prev};
      if ($charref_map->{$code}) {
        !!!cp (1015);
        !!!parse-error (type => 'invalid character reference',
                        text => (sprintf 'U+%04X', $code),
                        line => $l, column => $c);
        $code = $charref_map->{$code};
      } elsif ($code > 0x10FFFF) {
        !!!cp (1016);
        !!!parse-error (type => 'invalid character reference',
                        text => (sprintf 'U-%08X', $code),
                        line => $l, column => $c);
        $code = 0xFFFD;
      }

      if ($self->{prev_state} == DATA_STATE) {
        !!!cp (992);
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => CHARACTER_TOKEN, data => chr $code,
                  has_reference => 1,
                  line => $l, column => $c,
                 });
        redo A;
      } else {
        !!!cp (991);
        $self->{ca}->{value} .= chr $code;
        $self->{ca}->{has_reference} = 1;
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == HEXREF_X_STATE) {
      if ((0x0030 <= $self->{nc} and $self->{nc} <= 0x0039) or
          (0x0041 <= $self->{nc} and $self->{nc} <= 0x0046) or
          (0x0061 <= $self->{nc} and $self->{nc} <= 0x0066)) {
        # 0..9, A..F, a..f
        !!!cp (990);
        $self->{state} = HEXREF_HEX_STATE;
        $self->{kwd} = 0;
        ## Reconsume.
        redo A;
      } else {
        !!!parse-error (type => 'bare hcro',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 2);

        ## NOTE: According to the spec algorithm, nothing is returned,
        ## and then "&#" followed by "X" or "x" is appended to the parent
        ## element or the attribute value in the later processing.

        if ($self->{prev_state} == DATA_STATE) {
          !!!cp (1005);
          $self->{state} = $self->{prev_state};
          $self->{s_kwd} = '';
          ## Reconsume.
          !!!emit ({type => CHARACTER_TOKEN,
                    data => '&' . $self->{kwd},
                    line => $self->{line_prev},
                    column => $self->{column_prev} - length $self->{kwd},
                   });
          redo A;
        } else {
          !!!cp (989);
          $self->{ca}->{value} .= '&' . $self->{kwd};
          $self->{state} = $self->{prev_state};
          $self->{s_kwd} = '';
          ## Reconsume.
          redo A;
        }
      }
    } elsif ($self->{state} == HEXREF_HEX_STATE) {
      if (0x0030 <= $self->{nc} and $self->{nc} <= 0x0039) {
        # 0..9
        !!!cp (1002);
        $self->{kwd} *= 0x10;
        $self->{kwd} += $self->{nc} - 0x0030;
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif (0x0061 <= $self->{nc} and
               $self->{nc} <= 0x0066) { # a..f
        !!!cp (1003);
        $self->{kwd} *= 0x10;
        $self->{kwd} += $self->{nc} - 0x0060 + 9;
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif (0x0041 <= $self->{nc} and
               $self->{nc} <= 0x0046) { # A..F
        !!!cp (1004);
        $self->{kwd} *= 0x10;
        $self->{kwd} += $self->{nc} - 0x0040 + 9;
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003B) { # ;
        !!!cp (1006);
        !!!next-input-character;
        #
      } else {
        !!!cp (1007);
        !!!parse-error (type => 'no refc',
                        line => $self->{line},
                        column => $self->{column});
        ## Reconsume.
        #
      }

      my $code = $self->{kwd};
      my $l = $self->{line_prev};
      my $c = $self->{column_prev};
      if ($charref_map->{$code}) {
        !!!cp (1008);
        !!!parse-error (type => 'invalid character reference',
                        text => (sprintf 'U+%04X', $code),
                        line => $l, column => $c);
        $code = $charref_map->{$code};
      } elsif ($code > 0x10FFFF) {
        !!!cp (1009);
        !!!parse-error (type => 'invalid character reference',
                        text => (sprintf 'U-%08X', $code),
                        line => $l, column => $c);
        $code = 0xFFFD;
      }

      if ($self->{prev_state} == DATA_STATE) {
        !!!cp (988);
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => CHARACTER_TOKEN, data => chr $code,
                  has_reference => 1,
                  line => $l, column => $c,
                 });
        redo A;
      } else {
        !!!cp (987);
        $self->{ca}->{value} .= chr $code;
        $self->{ca}->{has_reference} = 1;
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == ENTITY_NAME_STATE) {
      if (length $self->{kwd} < 30 and
          ## NOTE: Some number greater than the maximum length of entity name
          ((0x0041 <= $self->{nc} and # a
            $self->{nc} <= 0x005A) or # x
           (0x0061 <= $self->{nc} and # a
            $self->{nc} <= 0x007A) or # z
           (0x0030 <= $self->{nc} and # 0
            $self->{nc} <= 0x0039) or # 9
           $self->{nc} == 0x003B)) { # ;
        our $EntityChar;
        $self->{kwd} .= chr $self->{nc};
        if (defined $EntityChar->{$self->{kwd}}) {
          if ($self->{nc} == 0x003B) { # ;
            !!!cp (1020);
            $self->{entity__value} = $EntityChar->{$self->{kwd}};
            $self->{entity__match} = 1;
            !!!next-input-character;
            #
          } else {
            !!!cp (1021);
            $self->{entity__value} = $EntityChar->{$self->{kwd}};
            $self->{entity__match} = -1;
            ## Stay in the state.
            !!!next-input-character;
            redo A;
          }
        } else {
          !!!cp (1022);
          $self->{entity__value} .= chr $self->{nc};
          $self->{entity__match} *= 2;
          ## Stay in the state.
          !!!next-input-character;
          redo A;
        }
      }

      my $data;
      my $has_ref;
      if ($self->{entity__match} > 0) {
        !!!cp (1023);
        $data = $self->{entity__value};
        $has_ref = 1;
        #
      } elsif ($self->{entity__match} < 0) {
        !!!parse-error (type => 'no refc');
        if ($self->{prev_state} != DATA_STATE and # in attribute
            $self->{entity__match} < -1) {
          !!!cp (1024);
          $data = '&' . $self->{kwd};
          #
        } else {
          !!!cp (1025);
          $data = $self->{entity__value};
          $has_ref = 1;
          #
        }
      } else {
        !!!cp (1026);
        !!!parse-error (type => 'bare ero',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - length $self->{kwd});
        $data = '&' . $self->{kwd};
        #
      }
  
      ## NOTE: In these cases, when a character reference is found,
      ## it is consumed and a character token is returned, or, otherwise,
      ## nothing is consumed and returned, according to the spec algorithm.
      ## In this implementation, anything that has been examined by the
      ## tokenizer is appended to the parent element or the attribute value
      ## as string, either literal string when no character reference or
      ## entity-replaced string otherwise, in this stage, since any characters
      ## that would not be consumed are appended in the data state or in an
      ## appropriate attribute value state anyway.
 
      if ($self->{prev_state} == DATA_STATE) {
        !!!cp (986);
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => CHARACTER_TOKEN,
                  data => $data,
                  has_reference => $has_ref,
                  line => $self->{line_prev},
                  column => $self->{column_prev} + 1 - length $self->{kwd},
                 });
        redo A;
      } else {
        !!!cp (985);
        $self->{ca}->{value} .= $data;
        $self->{ca}->{has_reference} = 1 if $has_ref;
        $self->{state} = $self->{prev_state};
        $self->{s_kwd} = '';
        ## Reconsume.
        redo A;
      }

    ## XML-only states

    } elsif ($self->{state} == PI_STATE) {
      ## XML5: "Pi state" and "DOCTYPE pi state".

      if ($is_space->{$self->{nc}} or
          $self->{nc} == 0x003F or # ?
          $self->{nc} == -1) {
        ## XML5: U+003F: "pi state": Same as "Anything else"; "DOCTYPE
        ## pi state": Switch to the "DOCTYPE pi after state".  EOF:
        ## "DOCTYPE pi state": Parse error, switch to the "data
        ## state".
        !!!parse-error (type => 'bare pio', ## TODO: type
                        line => $self->{line_prev},
                        column => $self->{column_prev}
                            - 1 * ($self->{nc} != -1));
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN,
                       data => '?',
                       line => $self->{line_prev},
                       column => $self->{column_prev}
                           - 1 * ($self->{nc} != -1),
                      };
        redo A;
      } else {
        ## XML5: "DOCTYPE pi state": Stay in the state.
        $self->{ct} = {type => PI_TOKEN,
                       target => chr $self->{nc},
                       data => '',
                       line => $self->{line_prev},
                       column => $self->{column_prev} - 1,
                      };
        $self->{state} = PI_TARGET_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == PI_TARGET_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = PI_TARGET_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'no pic'); ## TODO: type
        if ($self->{in_subset}) {
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## Reconsume.
        !!!emit ($self->{ct}); # pi
        redo A;
      } elsif ($self->{nc} == 0x003F) { # ?
        $self->{state} = PI_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } else {
        ## XML5: typo ("tag name" -> "target")
        $self->{ct}->{target} .= chr $self->{nc}; # pi
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == PI_TARGET_AFTER_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } else {
        $self->{state} = PI_DATA_STATE;
        ## Reprocess.
        redo A;
      }
    } elsif ($self->{state} == PI_DATA_STATE) {
      if ($self->{nc} == 0x003F) { # ?
        $self->{state} = PI_DATA_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'no pic'); ## TODO: type
        if ($self->{in_subset}) {
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state"
        } else {
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        ## Reprocess.
        !!!emit ($self->{ct}); # pi
        redo A;
      } else {
        $self->{ct}->{data} .= chr $self->{nc}; # pi
        $self->{read_until}->($self->{ct}->{data}, q[?],
                              length $self->{ct}->{data});
        ## Stay in the state.
        !!!next-input-character;
        ## Reprocess.
        redo A;
      }
    } elsif ($self->{state} == PI_AFTER_STATE) {
      ## XML5: Part of "Pi after state".

      if ($self->{nc} == 0x003E) { # >
        if ($self->{in_subset}) {
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;
        !!!emit ($self->{ct}); # pi
        redo A;
      } elsif ($self->{nc} == 0x003F) { # ?
        !!!parse-error (type => 'no s after target', ## TODO: type
                        line => $self->{line_prev},
                        column => $self->{column_prev}); ## XML5: no error
        $self->{ct}->{data} .= '?';
        $self->{state} = PI_DATA_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'no s after target', ## TODO: type
                        line => $self->{line_prev},
                        column => $self->{column_prev}
                            + 1 * ($self->{nc} == -1)); ## XML5: no error
        $self->{ct}->{data} .= '?'; ## XML5: not appended
        $self->{state} = PI_DATA_STATE;
        ## Reprocess.
        redo A;
      }
    } elsif ($self->{state} == PI_DATA_AFTER_STATE) {
      ## XML5: Same as "pi after state" and "DOCTYPE pi after state".

      if ($self->{nc} == 0x003E) { # >
        if ($self->{in_subset}) {
          $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        } else {
          $self->{state} = DATA_STATE;
          $self->{s_kwd} = '';
        }
        !!!next-input-character;
        !!!emit ($self->{ct}); # pi
        redo A;
      } elsif ($self->{nc} == 0x003F) { # ?
        $self->{ct}->{data} .= '?';
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } else {
        $self->{ct}->{data} .= '?'; ## XML5: not appended
        $self->{state} = PI_DATA_STATE;
        ## Reprocess.
        redo A;
      }

    } elsif ($self->{state} == DOCTYPE_INTERNAL_SUBSET_STATE) {
      if ($self->{nc} == 0x003C) { # <
        $self->{state} = DOCTYPE_TAG_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0025) { # %
        ## XML5: Not defined yet.

        ## TODO:
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x005D) { # ]
        delete $self->{in_subset};
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed internal subset'); ## TODO: type
        delete $self->{in_subset};
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => END_OF_DOCTYPE_TOKEN});
        redo A;
      } else {
        unless ($self->{internal_subset_tainted}) {
          ## XML5: No parse error.
          !!!parse-error (type => 'string in internal subset');
          $self->{internal_subset_tainted} = 1;
        }
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_INTERNAL_SUBSET_AFTER_STATE) {
      if ($self->{nc} == 0x003E) { # >
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;
        !!!emit ({type => END_OF_DOCTYPE_TOKEN});
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed DOCTYPE');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => END_OF_DOCTYPE_TOKEN});
        redo A;
      } else {
        ## XML5: No parse error and stay in the state.
        !!!parse-error (type => 'string after internal subset'); ## TODO: type

        $self->{state} = BOGUS_DOCTYPE_INTERNAL_SUBSET_AFTER_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == BOGUS_DOCTYPE_INTERNAL_SUBSET_AFTER_STATE) {
      if ($self->{nc} == 0x003E) { # >
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        !!!next-input-character;
        !!!emit ({type => END_OF_DOCTYPE_TOKEN});
        redo A;
      } elsif ($self->{nc} == -1) {
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        !!!emit ({type => END_OF_DOCTYPE_TOKEN});
        redo A;
      } else {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_TAG_STATE) {
      if ($self->{nc} == 0x0021) { # !
        $self->{state} = DOCTYPE_MARKUP_DECLARATION_OPEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003F) { # ?
        $self->{state} = PI_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'bare stago');
        $self->{state} = DATA_STATE;
        $self->{s_kwd} = '';
        ## Reconsume.
        redo A;
      } else {
        !!!parse-error (type => 'bare stago', ## XML5: Not a parse error.
                        line => $self->{line_prev},
                        column => $self->{column_prev});
        $self->{state} = BOGUS_COMMENT_STATE;
        $self->{ct} = {type => COMMENT_TOKEN,
                       data => '',
                      }; ## NOTE: Will be discarded.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_MARKUP_DECLARATION_OPEN_STATE) {
      ## XML5: "DOCTYPE markup declaration state".
      
      if ($self->{nc} == 0x002D) { # -
        $self->{state} = MD_HYPHEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0045) { # E
        $self->{state} = MD_E_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0041) { # A
        $self->{state} = MD_ATTLIST_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x004E) { # N
        $self->{state} = MD_NOTATION_STATE;
        $self->{kwd} = chr $self->{nc};
        !!!next-input-character;
        redo A;
      } else {
        #
      }
      
      ## XML5: No parse error.
      !!!parse-error (type => 'bogus comment',
                      line => $self->{line_prev},
                      column => $self->{column_prev} - 1);
      ## Reconsume.
      $self->{state} = BOGUS_COMMENT_STATE;
      $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded.
      redo A;
    } elsif ($self->{state} == MD_E_STATE) {
      if ($self->{nc} == 0x004E) { # N
        $self->{state} = MD_ENTITY_STATE;
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x004C) { # L
        ## XML5: <!ELEMENT> not supported.
        $self->{state} = MD_ELEMENT_STATE;
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } else {
        ## XML5: No parse error.
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 2
                            + 1 * ($self->{nc} == -1));
        ## Reconsume.
        $self->{state} = BOGUS_COMMENT_STATE;
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        redo A;
      }
    } elsif ($self->{state} == MD_ENTITY_STATE) {
      if ($self->{nc} == {
            'EN' => 0x0054, # T
            'ENT' => 0x0049, # I
            'ENTI' => 0x0054, # T
          }->{$self->{kwd}}) {
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{kwd} eq 'ENTIT' and
               $self->{nc} == 0x0059) { # Y
        $self->{ct} = {type => GENERAL_ENTITY_TOKEN, name => '', text => '',
                       line => $self->{line_prev},
                       column => $self->{column_prev} - 6};
        $self->{state} = DOCTYPE_MD_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1
                            - (length $self->{kwd})
                            + 1 * ($self->{nc} == -1));
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        redo A;
      }
    } elsif ($self->{state} == MD_ELEMENT_STATE) {
      if ($self->{nc} == {
            'EL' => 0x0045, # E
            'ELE' => 0x004D, # M
            'ELEM' => 0x0045, # E
            'ELEME' => 0x004E, # N
          }->{$self->{kwd}}) {
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{kwd} eq 'ELEMEN' and
               $self->{nc} == 0x0054) { # T
        $self->{ct} = {type => ELEMENT_TOKEN, name => '',
                       line => $self->{line_prev},
                       column => $self->{column_prev} - 6};
        $self->{state} = DOCTYPE_MD_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1
                            - (length $self->{kwd})
                            + 1 * ($self->{nc} == -1));
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        redo A;
      }
    } elsif ($self->{state} == MD_ATTLIST_STATE) {
      if ($self->{nc} == {
            'A' => 0x0054, # T
            'AT' => 0x0054, # T
            'ATT' => 0x004C, # L
            'ATTL' => 0x0049, # I
            'ATTLI' => 0x0053, # S
          }->{$self->{kwd}}) {
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{kwd} eq 'ATTLIS' and
               $self->{nc} == 0x0054) { # T
        $self->{ct} = {type => ATTLIST_TOKEN, name => '',
                       attrdefs => [],
                       line => $self->{line_prev},
                       column => $self->{column_prev} - 6};
        $self->{state} = DOCTYPE_MD_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1
                             - (length $self->{kwd})
                             + 1 * ($self->{nc} == -1));
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        redo A;
      }
    } elsif ($self->{state} == MD_NOTATION_STATE) {
      if ($self->{nc} == {
            'N' => 0x004F, # O
            'NO' => 0x0054, # T
            'NOT' => 0x0041, # A
            'NOTA' => 0x0054, # T
            'NOTAT' => 0x0049, # I
            'NOTATI' => 0x004F, # O
          }->{$self->{kwd}}) {
        ## Stay in the state.
        $self->{kwd} .= chr $self->{nc};
        !!!next-input-character;
        redo A;
      } elsif ($self->{kwd} eq 'NOTATIO' and
               $self->{nc} == 0x004E) { # N
        $self->{ct} = {type => NOTATION_TOKEN, name => '',
                       line => $self->{line_prev},
                       column => $self->{column_prev} - 6};
        $self->{state} = DOCTYPE_MD_STATE;
        !!!next-input-character;
        redo A;
      } else {
        !!!parse-error (type => 'bogus comment',
                        line => $self->{line_prev},
                        column => $self->{column_prev} - 1
                            - (length $self->{kwd})
                            + 1 * ($self->{nc} == -1));
        $self->{state} = BOGUS_COMMENT_STATE;
        ## Reconsume.
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_MD_STATE) {
      ## XML5: "DOCTYPE ENTITY state", "DOCTYPE ATTLIST state", and
      ## "DOCTYPE NOTATION state".

      if ($is_space->{$self->{nc}}) {
        ## XML5: [NOTATION] Switch to the "DOCTYPE NOTATION identifier state".
        $self->{state} = BEFORE_MD_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{ct}->{type} == GENERAL_ENTITY_TOKEN and 
               $self->{nc} == 0x0025) { # %
        ## XML5: Switch to the "DOCTYPE bogus comment state".
        !!!parse-error (type => 'no space before md name'); ## TODO: type
        $self->{state} = DOCTYPE_ENTITY_PARAMETER_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        ## Reconsume.
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Switch to the "DOCTYPE bogus comment state".
        !!!parse-error (type => 'no md name'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        redo A;
      } else {
        ## XML5: Switch to the "DOCTYPE bogus comment state".
        !!!parse-error (type => 'no space before md name'); ## TODO: type
        $self->{state} = BEFORE_MD_NAME_STATE;
        redo A;
      }
    } elsif ($self->{state} == BEFORE_MD_NAME_STATE) {
      ## XML5: "DOCTYPE ENTITY parameter state", "DOCTYPE ENTITY type
      ## before state", "DOCTYPE ATTLIST name before state".

      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{ct}->{type} == GENERAL_ENTITY_TOKEN and 
               $self->{nc} == 0x0025) { # %
        $self->{state} = DOCTYPE_ENTITY_PARAMETER_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "Anything else".
        !!!parse-error (type => 'no md name'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        ## Reconsume.
        redo A;
      } else {
        ## XML5: [ATTLIST] Not defined yet.
        $self->{ct}->{name} .= chr $self->{nc};
        $self->{state} = MD_NAME_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ENTITY_PARAMETER_BEFORE_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## XML5: Switch to the "DOCTYPE ENTITY parameter state".
        $self->{ct}->{type} = PARAMETER_ENTITY_TOKEN;
        $self->{state} = BEFORE_MD_NAME_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "Anything else".
        !!!parse-error (type => 'no md name'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed md');
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        ## Reconsume.
        redo A;
      } else {
        ## XML5: No parse error.
        !!!parse-error (type => 'no space after ENTITY percent'); ## TODO: type
        $self->{state} = BOGUS_COMMENT_STATE;
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == MD_NAME_STATE) {
      ## XML5: "DOCTYPE ENTITY name state" and "DOCTYPE ATTLIST name state".
      
      if ($is_space->{$self->{nc}}) {
        ## TODO:
        $self->{state} = DOCTYPE_ATTLIST_NAME_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        if ($self->{ct}->{type} == ATTLIST_TOKEN) {
          #
        } else {
          !!!parse-error (type => 'no md body'); ## TODO: type
        }
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ELEMENT/ENTITY/ATTLIST/NOTATION
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: [ATTLIST] No parse error.
        !!!parse-error (type => 'unclosed md');
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        ## Reconsume.
        !!!emit ($self->{ct}); # ELEMENT/ENTITY/ATTLIST/NOTATION
        redo A;
      } else {
        ## XML5: [ATTLIST] Not defined yet.
        $self->{ct}->{name} .= chr $self->{nc};
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_NAME_AFTER_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!emit ($self->{ct});
        redo A;
      } else {
        ## XML5: Not defined yet.
        $self->{ca} = {name => chr ($self->{nc}), # attrdef
                       tokens => [],
                       line => $self->{line}, column => $self->{column}};
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_NAME_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_NAME_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_NAME_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no attr type'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == 0x0028) { # (
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before paren'); ## TODO: type
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } else {
        ## XML5: Not defined yet.
        $self->{ca}->{name} .= chr $self->{nc};
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_NAME_AFTER_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no attr type'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == 0x0028) { # (
        ## XML5: Same as "anything else".
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        ## XML5: Not defined yet.
        $self->{ca}->{type} = chr $self->{nc};
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0023) { # #
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no attr default'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == 0x0028) { # (
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before paren'); ## TODO: type
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        ## XML5: Not defined yet.
        $self->{ca}->{type} .= chr $self->{nc};
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_TYPE_AFTER_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0028) { # (
        ## XML5: Same as "anything else".
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0023) { # #
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        ## XML5: Same as "anything else".
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        ## XML5: Same as "anything else".
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no attr default'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        ## XML5: Switch to the "DOCTYPE bogus comment state".
        !!!parse-error (type => 'unquoted attr value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == BEFORE_ALLOWED_TOKEN_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x007C) { # |
        !!!parse-error (type => 'empty allowed token'); ## TODO: type
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0029) { # )
        !!!parse-error (type => 'empty allowed token'); ## TODO: type
        $self->{state} = AFTER_ALLOWED_TOKENS_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'unclosed allowed tokens'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        push @{$self->{ca}->{tokens}}, chr $self->{nc};
        $self->{state} = ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == ALLOWED_TOKEN_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = AFTER_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x007C) { # |
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0029) { # )
        $self->{state} = AFTER_ALLOWED_TOKENS_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'unclosed allowed tokens'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        $self->{ca}->{tokens}->[-1] .= chr $self->{nc};
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_ALLOWED_TOKEN_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x007C) { # |
        $self->{state} = BEFORE_ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0029) { # )
        $self->{state} = AFTER_ALLOWED_TOKENS_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'unclosed allowed tokens'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        !!!parse-error (type => 'space in allowed token', ## TODO: type
                        line => $self->{line_prev},
                        column => $self->{column_prev});
        $self->{ca}->{tokens}->[-1] .= ' ' . chr $self->{nc};
        $self->{state} = ALLOWED_TOKEN_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == AFTER_ALLOWED_TOKENS_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = BEFORE_ATTR_DEFAULT_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0023) { # #
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'no attr default'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        !!!parse-error (type => 'unquoted attr value'); ## TODO: type
        $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == BEFORE_ATTR_DEFAULT_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0023) { # #
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        !!!parse-error (type => 'no attr default'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        !!!parse-error (type => 'unquoted attr value'); ## TODO: type
        $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_BEFORE_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## XML5: No parse error.
        !!!parse-error (type => 'no default type'); ## TODO: type
        $self->{state} = BOGUS_COMMENT_STATE;
        $self->{ct} = {type => COMMENT_TOKEN, data => ''}; ## Will be discarded
        ## Reconsume.
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        ## XML5: Same as "anything else".
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        ## XML5: Same as "anything else".
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no attr default'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        $self->{ca}->{default} = chr $self->{nc};
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_STATE;
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_STATE) {
      if ($is_space->{$self->{nc}}) {
        $self->{state} = DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_AFTER_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        ## XML5: Same as "anything else".
        !!!parse-error (type => 'no space before default value'); ## TODO: type
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        ## XML5: Same as "anything else".
        push @{$self->{ct}->{attrdefs}}, $self->{ca};
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        push @{$self->{ct}->{attrdefs}}, $self->{ca};
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        $self->{ca}->{default} .= chr $self->{nc};
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      }
    } elsif ($self->{state} == DOCTYPE_ATTLIST_ATTRIBUTE_DECLARATION_AFTER_STATE) {
      if ($is_space->{$self->{nc}}) {
        ## Stay in the state.
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0022) { # "
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_DOUBLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x0027) { # '
        $self->{ca}->{value} = '';
        $self->{state} = ATTRIBUTE_VALUE_SINGLE_QUOTED_STATE;
        !!!next-input-character;
        redo A;
      } elsif ($self->{nc} == 0x003E) { # >
        push @{$self->{ct}->{attrdefs}}, $self->{ca};
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE;
        !!!next-input-character;
        !!!emit ($self->{ct}); # ATTLIST
        redo A;
      } elsif ($self->{nc} == -1) {
        ## XML5: No parse error.
        !!!parse-error (type => 'unclosed md'); ## TODO: type
        push @{$self->{ct}->{attrdefs}}, $self->{ca};
        $self->{state} = DOCTYPE_INTERNAL_SUBSET_STATE; ## XML5: "Data state".
        !!!next-input-character;
        !!!emit ($self->{ct});
        redo A;
      } else {
        ## XML5: Not defined yet.
        if ($self->{ca}->{default} eq 'FIXED') {
          $self->{state} = ATTRIBUTE_VALUE_UNQUOTED_STATE;
        } else {
          push @{$self->{ct}->{attrdefs}}, $self->{ca};
          $self->{state} = DOCTYPE_ATTLIST_NAME_AFTER_STATE;
        }
        ## Reconsume.
        redo A;
      }
    } elsif ($self->{state} == AFTER_ATTLIST_ATTR_VALUE_QUOTED_STATE) {
      if ($is_space->{$self->{nc}} or
          $self->{nc} == -1 or
          $self->{nc} == 0x003E) { # >
        $self->{state} = DOCTYPE_ATTLIST_NAME_AFTER_STATE;
        ## Reconsume.
        redo A;
      } else {
        !!!parse-error (type => 'no space before attr name'); ## TODO: type
        $self->{state} = DOCTYPE_ATTLIST_NAME_AFTER_STATE;
        ## Reconsume.
        redo A;
      }      
    } else {
      die "$0: $self->{state}: Unknown state";
    }
  } # A   

  die "$0: _get_next_token: unexpected case";
} # _get_next_token

1;
## $Date: 2008/10/17 07:14:29 $
                                