#!/usr/local/bin/perl

# refcl - reference cleaner
#   Replaces HTML character entity references with New-JIS (JIS X 0213)
#   characters. e.g., € is replaced with a euro sign.  EUC only.

#   Special notation: &jis-[men]-[ku]-[ten];
#       stands for JIS X 0213 code point.
#       e.g., &jis-1-6-57; is replaced with a "final sigma".

# This script is based on the New-JIS code table X0213-p.csv 
# provided by JCS/WG2 and W3C's HTML 4.0 spec.

# Feel free to copy and modify.
# No warranty.

%map = (
  nbsp => a9a2,
  iexcl => a9a3,
  curren => a9a4,
  brvbar => a9a5,
  copy => a9a6,
  ordf => a9a7,
  laquo => a9a8,
  shy => a9a9,
  reg => a9aa,
  macr => a9ab,
  sup2 => a9ac,
  sup3 => a9ad,
  middot => a9ae,
  cedil => a9af,
  sup1 => a9b0,
  ordm => a9b1,
  raquo => a9b2,
  frac14 => a9b3,
  frac12 => a9b4,
  frac34 => a9b5,
  iquest => a9b6,
  Agrave => a9b7,
  Aacute => a9b8,
  Acirc => a9b9,
  Atilde => a9ba,
  Auml => a9bb,
  Aring => a9bc,
  AElig => a9bd,
  Ccedil => a9be,
  Egrave => a9bf,
  Eacute => a9c0,
  Ecirc => a9c1,
  Euml => a9c2,
  Igrave => a9c3,
  Iacute => a9c4,
  Icirc => a9c5,
  Iuml => a9c6,
  ETH => a9c7,
  Ntilde => a9c8,
  Ograve => a9c9,
  Oacute => a9ca,
  Ocirc => a9cb,
  Otilde => a9cc,
  Ouml => a9cd,
  Oslash => a9ce,
  Ugrave => a9cf,
  Uacute => a9d0,
  Ucirc => a9d1,
  Uuml => a9d2,
  Yacute => a9d3,
  THORN => a9d4,
  szlig => a9d5,
  agrave => a9d6,
  aacute => a9d7,
  acirc => a9d8,
  atilde => a9d9,
  auml => a9da,
  aring => a9db,
  aelig => a9dc,
  ccedil => a9dd,
  egrave => a9de,
  eacute => a9df,
  ecirc => a9e0,
  euml => a9e1,
  igrave => a9e2,
  iacute => a9e3,
  icirc => a9e4,
  iuml => a9e5,
  eth => a9e6,
  ntilde => a9e7,
  ograve => a9e8,
  oacute => a9e9,
  ocirc => a9ea,
  otilde => a9eb,
  ouml => a9ec,
  oslash => a9ed,
  ugrave => a9ee,
  uacute => a9ef,
  ucirc => a9f0,
  uuml => a9f1,
  yacute => a9f2,
  thorn => a9f3,
  yuml => a9f4,
  OElig => abab,
  oelig => abaa,
  Scaron => aaa6,
  scaron => aab2,
  ndash => a3fc,
  euro => a9a1,
  sigmaf => a6d9,
  bull => a3c0,
  alefsym => a3dc,
  harr => a2f1,
  empty => a2c7,
  notin => a2c6,
  cong => a2ed,
  asymp => a2ee,
  nsub => a2c2,
  oplus => a2d1,
  otimes => a2d3,
  spades => a6ba,
  clubs => a6c0,
  hearts => a6be,
  diams => a6bc
    );

while(<>) {
    while (/\&([a-zA-Z0-9-.]+);/) {
	print $`;
	$matched = $&;
	$_ = $';
	if ($matched =~ /jis\-([12])\-(\d\d?)\-(\d\d?)/) { # special notation
	    $euc = pack "CC", 160 + $2, 160 + $3;
	    $matched = (($1 == 2) ? (pack H2, "8f") : "") . $euc;
	} else {
	    $euc = $map{"$1"};
	    if ($euc ne "") {
		$subst = pack H4, $euc;
		$matched =~ s/\&([a-zA-Z0-9-.]+);/$subst/;
	    }
	}
	print $matched;
    }
    print $_;
}