# This does what Unicode::Escape::escape does, except a lot less.
# It assumes the input is utf8 encoded.
# It assumes the input is defined.
# Invalid utf8 is deleted silently.

sub utf8_escape {
    # Short circuit if it's all seven bit ASCII.
    return $_[0] unless $_[0] =~ /[\x80-\xff]/;

    my $s = shift;

    $s =~ s{ ([\xc2-\xdf]) ([\x80-\xbf]) }{
        '\\u' .
        sprintf( '%04x',
              ( ( 0b00011111 & ord $1 ) << 6 )
            |   ( 0b00111111 & ord $2 )
        )
    }exmsg;
    $s =~ s{ ([\xe0-\xef]) ([\x80-\xbf]) ([\x80-\xbf]) }{
        '\\u' . 
        sprintf( '%04x',
              ( ( 0b00001111 & ord $1 ) << 12 )
            | ( ( 0b00111111 & ord $2 ) <<  6 )
            |   ( 0b00111111 & ord $3 )
        )
    }exmsg;

    # valid utf8 that can't be encoded in \uXXXX
    $s =~ s{ [\xf0-\xf4] [\x80-\xbf]{3} }{\\ufffd}xmsg;

    # invalid utf8
    $s =~ tr/\x80-\xff//d;

    return $s;
}