#!/usr/bin/perl -s

our ($f,$force);
$f = $force if $force;

use strict;
my $charset;

for(@ARGV){
  $charset="?";
  next if /.bak$/;
  my $tipo = `file -b -i '$_'`;
  chomp $tipo;
  if($tipo =~ /(\S+);\s+charset=(\S+)/){$tipo=$1; $charset=lc($2)};

     if($charset eq "utf-8")   {  }
  elsif($charset eq "binary")   {  }
  elsif($charset eq "us-ascii")   {  }
  elsif($tipo eq "application/postscript"){ }
  elsif(($tipo eq "application/xml" or $tipo eq "text/xml") and ($charset eq "iso-8859"  or  $charset eq "iso-8859-1")){
    rename($_ ,"$_.bak");
##    system ("iconv -f CP1252 -t utf-8 '$_.bak' > '$_.aux'\n")  ;
    JJsystem ("xmllint --encode utf-8 '$_.bak' > '$_'\n")  }
  elsif($tipo eq "text/x-tex" and ($charset eq "iso-8859"  or  $charset eq "iso-8859-1")){
    rename($_ ,"$_.bak");
    JJsystem ("iconv -f CP1252 -t utf-8 '$_.bak' | sed 's/latin1\]\{inputenc\}/utf8]{inputenc}/g' > '$_'") }
  elsif($tipo eq "application/xml" or $tipo eq "text/xml"){
    rename($_ ,"$_.bak");
    JJsystem ("xmllint --encode utf-8 '$_.bak' > '$_'\n")  }
  elsif($charset eq "non-iso" or $charset eq "unknown-8bit" ) { 
    rename($_ ,"$_.bak");
    JJsystem ("iconv -f CP1252 -t utf-8 '$_.bak' > '$_'\n")  }
  elsif($charset eq "utf-16le"){
    rename($_ ,"$_.bak");
    JJsystem ("iconv -f $charset  -t utf-8 '$_.bak' > '$_'\n")  }
  elsif($charset eq "iso-8859"  or  $charset eq "iso-8859-1"){
    rename($_ ,"$_.bak");
    JJsystem ("iconv -f latin1 -t utf-8 '$_.bak' > '$_'\n")  }
  else                      { print "socorro: '$tipo:$charset' '$_'\n"; }
}

sub JJsystem{ my ($com)=@_;
  if(system($com)!=0){
     print stderr "Error in $com:$!;\nReverting...\n";
     rename("$_.bak",$_);
  }
}

__END__
dominio/Dominio_10_A.txt:           text/plain; charset=utf-8
dominio/Dominio_10_A.txt.bak:       text/plain; charset=iso-8859-1

text/plain; charset=us-ascii
application/xml; charset=unknown-8bit
application/xml; charset=utf-8
application/xml; charset=utf-8
application/xml; charset=iso-8859-1

__END__

=head1 NAME

ensureutf8 - convert text files to unicode utf8

=head1 SYNOPSIS

 ensureutf8 file*
    -f -force         continue when error detected error 

=head1 DESCRIPTION

Each file is converted to utf8 (original is saved with extension .bak).

This command uses "file" to obtain original encoding.

 * file -i -b ( to get type and charset)
 * for TXT-file:  ( iconv ...)
 * for XML-files: ( xmllint --encode utf-8 )

=head1 AUTHOR

J.Joao Almeida, jj@di.uminho.pt

=head1 SEE ALSO

perl(1).

=cut