#!/usr/bin/perl
use strict;
use warnings;

sub help {
  print "Replaces non-ASCII accented and other characters with HTML entities.\n";
  print "The non-ASCII characters replaced are, generally:\n";
  print "  aeiou with acute, circumflex, grave, ring above, tilde, or diaeresis;\n";
  print "  slashed O and o;\n";
  print "  ae and AE ligatures;\n";
  print "  c and C with cedilla;\n";
  print "  ess;\n";
  print "  n and N with tilde;\n";
  print "  single and double quotes.\n";
  print "Certain non-ASCII characters are not handled:\n";
  print "  ETH, THORN, Y acute, division, eth, thorn, times, y acute.\n";
  print "Input is taken from files named on the command line or from STDIN.\n";
  print "Options:\n";
  print "  -h, -help      Show this help.\n";
  print "  -e, -entities  List the HTML entities in the input.\n";
  print "  -n, -non-ASCII List the non-ASCII characters (by their entity names)\n";
  print "                   in the input [default].\n";
  print "  -r, -replace   Print the input with non-ASCII characters replaced by\n";
  print "                   entities.\n";
  print "  -w, -where     Show where each non-ASCII character or entity appeared.\n";
}

#print "\$\#ARGV: $#ARGV\n";

my $act = "n";  # e(ntities), n(on-ASCII), r(eplace)
my $showLocation = "";
my %found = ();
my $entity = "";
my $fileCount = 0;
my $arg = "";
my $fname = "";
my $changedLines = 0;
my $v = "y";

if (0 == scalar @ARGV) {
  &help;
  exit 0;
}

foreach $arg (@ARGV) {
  if ($arg =~ /^-/) {
    if    ($arg =~ /^-+h(elp)?$/) {
      &help;
      exit 0;
    }
    elsif ($arg =~ /^-+e(ntities)?$/) {
      $act = "e";
    }
    elsif ($arg =~ /^-+n(on-ASCII)?$/) {
      $act = "n";
    }
    elsif ($arg =~ /^-+r(eplace)?$/) {
      $act = "r";
    }
    elsif ($arg =~ /^-+w(here)?$/) {
      $showLocation = "y";
    }
    else {
      print "Unexpected option \"$arg\".\n";
      exit -1;
    }
  }
  else {
    $fname = $arg;
    if (open HIN, $fname) {
      &perFile(\*HIN, $fname);
      close HIN;
      $fileCount = 1 + $fileCount;
    }
    else {
      print "Unable to open input file \"$arg\".\n";
      exit -1;
    }
  }
}

if (0 == $fileCount) {
  &perFile(\*STDIN, "<stdin>");
}
# if (0 < $changedLines) {
#   print STDERR "Changed lines: $changedLines.\n";
# }

sub perFile {
  if    ($act eq "e") {
    %found = ();
    &showEntities(@_);
    &showFound;
  }
  elsif ($act eq "n") {
    %found = ();
    &showAccents(@_);
    &showFound;
  }
  elsif ($act eq "r") {
    &replaceAccents(@_);
  }
  else {
    die "Unexpected action \"$act\"";
  }
}

sub replaceAccents {
  my $IN    = $_[0];
  my $fname = $_[1];
  my $lineNumber = 0;
  if ($v) {  print STDERR "Replacing accents in \"$fname\":\n";  }
  while (<$IN>) {
    ++$lineNumber;
    my $line0 = $_;

    # Alphabetized by HTML character entity name.

    s/�/&AElig;/g;   # latin capital letter AE

    s/�/&Aacute;/g;  # latin capital letter A with acute
    s/�/&Acirc;/g;   # latin capital letter A with circumflex
    s/�/&Agrave;/g;  # latin capital letter A with grave
    s/�/&Aring;/g;   # latin capital letter A with ring above
    s/�/&Atilde;/g;  # latin capital letter A with tilde
    s/�/&Auml;/g;    # latin capital letter A with diaeresis

    s/�/&Ccedil;/g;  # latin capital letter C with cedilla

   #s/?/&ETH;/g;     # latin capital letter ETH, U+00D0 ISOlat1

    s/�/&Eacute;/g;  # latin capital letter E with acute
    s/�/&Ecirc;/g;   # latin capital letter E with circumflex
    s/�/&Egrave;/g;  # latin capital letter E with grave
    s/�/&Euml;/g;    # latin capital letter E with diaeresis

    s/�/&Iacute;/g;  # latin capital letter I with acute
    s/�/&Icirc;/g;   # latin capital letter I with circumflex
    s/�/&Igrave;/g;  # latin capital letter I with grave
    s/�/&Iuml;/g;    # latin capital letter I with diaeresis

    s/�/&Ntilde;/g;  # latin capital letter N with tilde

    s/�/&Oacute;/g;  # latin capital letter O with acute
    s/�/&Ocirc;/g;   # latin capital letter O with circumflex
    s/�/&Ograve;/g;  # latin capital letter O with grave
    s/�/&Oslash;/g;  # latin capital letter O with stroke
    s/�/&Otilde;/g;  # latin capital letter O with tilde
    s/�/&Ouml;/g;    # latin capital letter O with diaeresis

   #s/?/&THORN;/g;   # latin capital letter THORN

    s/�/&Uacute;/g;  # latin capital letter U with acute
    s/�/&Ucirc;/g;   # latin capital letter U with circumflex
    s/�/&Ugrave;/g;  # latin capital letter U with grave
    s/�/&Uuml;/g;    # latin capital letter U with diaeresis

   #s/?/&Yacute;/g;  # latin capital letter Y with acute

    s/�/&aacute;/g;  # latin small letter a with acute
    s/�/&acirc;/g;   # latin small letter a with circumflex

    s/�/&aelig;/g;   # latin small letter ae, U+00E6 ISOlat1

    s/�/&agrave;/g;  # latin small letter a with grave
    s/�/&aring;/g;   # latin small letter a with ring above
    s/�/&atilde;/g;  # latin small letter a with tilde
    s/�/&auml;/g;    # latin small letter a with diaeresis

    s/�/&ccedil;/g;  # latin small letter c with cedilla
   #s/�/&divide;/g;  # division sign, U+00F7 ISOnum

    s/�/&eacute;/g;  # latin small letter e with acute
    s/�/&ecirc;/g;   # latin small letter e with circumflex
    s/�/&egrave;/g;  # latin small letter e with grave

   #s/?/&eth;/g;     # latin small letter eth, U+00F0 ISOlat1

    s/�/&euml;/g;    # latin small letter e with diaeresis

    s/�/&iacute;/g;  # latin small letter i with acute
    s/�/&icirc;/g;   # latin small letter i with circumflex
    s/�/&igrave;/g;  # latin small letter i with grave
    s/�/&iuml;/g;    # latin small letter i with diaeresis

    s/�/&ntilde;/g;  # latin small letter n with tilde

    s/�/&oacute;/g;  # latin small letter o with acute
    s/�/&ocirc;/g;   # latin small letter o with circumflex
    s/�/&ograve;/g;  # latin small letter o with grave
    s/�/&oslash;/g;  # latin small letter o with stroke,
    s/�/&otilde;/g;  # latin small letter o with tilde
    s/�/&ouml;/g;    # latin small letter o with diaeresis

    s/�/&szlig;/g;   # latin small letter sharp s = ess-zed

   #s/?/&thorn;/g;   # latin small letter thorn

   #s/?/&times;/g;   # multiplication sign, U+00D7 ISOnum

    s/�/&uacute;/g;  # latin small letter u with acute
    s/�/&ucirc;/g;   # latin small letter u with circumflex
    s/�/&ugrave;/g;  # latin small letter u with grave
    s/�/&uuml;/g;    # latin small letter u with diaeresis

   #s/?/&yacute;/g;  # latin small letter y with acute
    s/�/&yuml;/g;    # latin small letter y with diaeresis

    s/�/&ldquo;/g;   # ``
    s/�/&rdquo;/g;   # ''
    s/�/&lsquo;/g;   # `
    s/�/&rsquo;/g;   # '

    s/�/&bull;/g;    # bullet

    s/�/&mdash;/g;   # m dash

    s/�/ /g;         # Firefox shows it as Eacute.

    if ($_ ne $line0) {
      ++$changedLines;
      if ($showLocation) {
        print STDERR "($fname:$lineNumber)\n";
      }
    }

    print $_;
  }
}

sub deaccent {
  s/�/AE/g;   # latin capital letter AE

  s/�/A/g;  # latin capital letter A with acute
  s/�/A/g;  # latin capital letter A with circumflex
  s/�/A/g;  # latin capital letter A with grave
  s/�/A/g;  # latin capital letter A with ring above
  s/�/A/g;  # latin capital letter A with tilde
  s/�/A/g;  # latin capital letter A with diaeresis

  s/�/C/g;  # latin capital letter C with cedilla

  s/�/E/g;  # latin capital letter E with acute
  s/�/E/g;  # latin capital letter E with circumflex
  s/�/E/g;  # latin capital letter E with grave
  s/�/E/g;  # latin capital letter E with diaeresis

  s/�/I/g;  # latin capital letter I with acute
  s/�/I/g;  # latin capital letter I with circumflex
  s/�/I/g;  # latin capital letter I with grave
  s/�/I/g;  # latin capital letter I with diaeresis

  s/�/N/g;  # latin capital letter N with tilde

  s/�/O/g;  # latin capital letter O with acute
  s/�/O/g;  # latin capital letter O with circumflex
  s/�/O/g;  # latin capital letter O with grave
  s/�/O/g;  # latin capital letter O with stroke
  s/�/O/g;  # latin capital letter O with tilde
  s/�/O/g;  # latin capital letter O with diaeresis

  s/�/U/g;  # latin capital letter U with acute
  s/�/U/g;  # latin capital letter U with circumflex
  s/�/U/g;  # latin capital letter U with grave
  s/�/U/g;  # latin capital letter U with diaeresis

  s/�/a/g;  # latin small letter a with acute
  s/�/a/g;  # latin small letter a with circumflex

  s/�/a/g;  # latin small letter ae, U+00E6 ISOlat1

  s/�/a/g;  # latin small letter a with grave
  s/�/a/g;  # latin small letter a with ring above
  s/�/a/g;  # latin small letter a with tilde
  s/�/a/g;  # latin small letter a with diaeresis

  s/�/c/g;  # latin small letter c with cedilla

  s/�/e/g;  # latin small letter e with acute
  s/�/e/g;  # latin small letter e with circumflex
  s/�/e/g;  # latin small letter e with grave

  s/�/e/g;  # latin small letter e with diaeresis

  s/�/i/g;  # latin small letter i with acute
  s/�/i/g;  # latin small letter i with circumflex
  s/�/i/g;  # latin small letter i with grave
  s/�/i/g;  # latin small letter i with diaeresis

  s/�/n/g;  # latin small letter n with tilde

  s/�/o/g;  # latin small letter o with acute
  s/�/o/g;  # latin small letter o with circumflex
  s/�/o/g;  # latin small letter o with grave
  s/�/o/g;  # latin small letter o with stroke,
  s/�/o/g;  # latin small letter o with tilde
  s/�/o/g;  # latin small letter o with diaeresis

  s/�/s/g;  # latin small letter sharp s = ess-zed

  s/�/u/g;  # latin small letter u with acute
  s/�/u/g;  # latin small letter u with circumflex
  s/�/u/g;  # latin small letter u with grave
  s/�/u/g;  # latin small letter u with diaeresis

  s/�/y/g;  # latin small letter y with diaeresis
}

sub showAccents {
  my $IN    = $_[0];
  my $fname = $_[1];
  if ($v) {  print STDERR "Showing accents in \"$fname\":\n";  }
  my $lineno = 0;
  while (<$IN>) {
    ++$lineno;
    my $line0 = $_;

    # Alphabetized by HTML character entity name.

    if (/�/) {  show($fname, $lineno, "AElig");  }

    if (/�/) {  show($fname, $lineno, "Aacute");  }
    if (/�/) {  show($fname, $lineno, "Acirc");  }
    if (/�/) {  show($fname, $lineno, "Agrave");  }
    if (/�/) {  show($fname, $lineno, "Aring");  }
    if (/�/) {  show($fname, $lineno, "Atilde");  }
    if (/�/) {  show($fname, $lineno, "Auml");  }

    if (/�/) {  show($fname, $lineno, "Ccedil");  }

   #if (/?/) {  show($fname, $lineno, "ETH");  }

    if (/�/) {  show($fname, $lineno, "Eacute");  }
    if (/�/) {  show($fname, $lineno, "Ecirc");  }
    if (/�/) {  show($fname, $lineno, "Egrave");  }
    if (/�/) {  show($fname, $lineno, "Euml");  }

    if (/�/) {  show($fname, $lineno, "Iacute");  }
    if (/�/) {  show($fname, $lineno, "Icirc");  }
    if (/�/) {  show($fname, $lineno, "Igrave");  }
    if (/�/) {  show($fname, $lineno, "Iuml");  }

    if (/�/) {  show($fname, $lineno, "Ntilde");  }

    if (/�/) {  show($fname, $lineno, "Oacute");  }
    if (/�/) {  show($fname, $lineno, "Ocirc");  }
    if (/�/) {  show($fname, $lineno, "Ograve");  }
    if (/�/) {  show($fname, $lineno, "Oslash");  }
    if (/�/) {  show($fname, $lineno, "Otilde");  }
    if (/�/) {  show($fname, $lineno, "Ouml");  }

   #if (/?/) {  show($fname, $lineno, "THORN");  }

    if (/�/) {  show($fname, $lineno, "Uacute");  }
    if (/�/) {  show($fname, $lineno, "Ucirc");  }
    if (/�/) {  show($fname, $lineno, "Ugrave");  }
    if (/�/) {  show($fname, $lineno, "Uuml");  }

   #if (/?/) {  show($fname, $lineno, "Yacute");  }

    if (/�/) {  show($fname, $lineno, "aacute");  }
    if (/�/) {  show($fname, $lineno, "acirc");  }

    if (/�/) {  show($fname, $lineno, "aelig");  }

    if (/�/) {  show($fname, $lineno, "agrave");  }
    if (/�/) {  show($fname, $lineno, "aring");  }
    if (/�/) {  show($fname, $lineno, "atilde");  }
    if (/�/) {  show($fname, $lineno, "auml");  }

    if (/�/) {  show($fname, $lineno, "ccedil");  }
   #if (/�/) {  show($fname, $lineno, "divide");  }

    if (/�/) {  show($fname, $lineno, "eacute");  }
    if (/�/) {  show($fname, $lineno, "ecirc");  }
    if (/�/) {  show($fname, $lineno, "egrave");  }

   #if (/?/) {  show($fname, $lineno, "eth");  }

    if (/�/) {  show($fname, $lineno, "euml");  }

    if (/�/) {  show($fname, $lineno, "iacute");  }
    if (/�/) {  show($fname, $lineno, "icirc");  }
    if (/�/) {  show($fname, $lineno, "igrave");  }
    if (/�/) {  show($fname, $lineno, "iuml");  }

    if (/�/) {  show($fname, $lineno, "ntilde");  }

    if (/�/) {  show($fname, $lineno, "oacute");  }
    if (/�/) {  show($fname, $lineno, "ocirc");  }
    if (/�/) {  show($fname, $lineno, "ograve");  }
    if (/�/) {  show($fname, $lineno, "oslash");  }
    if (/�/) {  show($fname, $lineno, "otilde");  }
    if (/�/) {  show($fname, $lineno, "ouml");  }

    if (/�/) {  show($fname, $lineno, "szlig");  }

   #if (/?/) {  show($fname, $lineno, "thorn");  }

   #if (/?/) {  show($fname, $lineno, "times");  }

    if (/�/) {  show($fname, $lineno, "uacute");  }
    if (/�/) {  show($fname, $lineno, "ucirc");  }
    if (/�/) {  show($fname, $lineno, "ugrave");  }
    if (/�/) {  show($fname, $lineno, "uuml");  }

   #if (/?/) {  show($fname, $lineno, "yacute");  }
    if (/�/) {  show($fname, $lineno, "yuml");  }

    if (/�/) {  show($fname, $lineno, "ldquo");  }
    if (/�/) {  show($fname, $lineno, "rdquo");  }
    if (/�/) {  show($fname, $lineno, "lsquo");  }
    if (/�/) {  show($fname, $lineno, "rsquo");  }

    if (/�/) {  show($fname, $lineno, "bull");  }
    if (/�/) {  show($fname, $lineno, "mdash");  }

  }
}

sub showEntities {
  my $IN    = $_[0];
  my $fname = $_[1];
  if ($v) {  print STDERR "Showing entities in \"$fname\":\n";  }
  my $lineno = 0;
  $entity = "";

  while (<$IN>) {
    ++$lineno;
    chomp;
    while (/&[A-Za-z#0-9]+;/) {
      #print "line:  $_";
      s/(.*&[A-Za-z#0-9]+;).*$/$1/;
      #print "without trailing characters:  $_";
      $entity = $_;
      #print "Has entity:  \"$entity\".\n";
      $entity =~ s/.*(&[A-Za-z#0-9]+;)$/$1/;
      #print "Entity:  \"$entity\".\n";
      &show($fname, $lineno, $entity);
      s/(.*)&[A-Za-z#0-9]+;$/$1/;
      #print "without last entity:  $_";
    }
  }
}


sub show {
  my $fname  = $_[0];
  my $lineno = $_[1];
  $entity    = $_[2];

  if ($showLocation) {
    print "($fname:$lineno)  '$entity'\n";
  }
  else {
    $found{$entity} = "($fname:$lineno)";
  }
}

sub showFound {
  if ($showLocation) {
    # do nothing
  }
  else {
    my $count = scalar keys %found;
    if ($count) {
      foreach $entity (sort keys %found) {
        print "$entity\n";
      }
    }
    print STDERR "$count found\n";
  }
}