Exchanging RTF with MacOSX

[Andreas Höschler]

Hi Wolfgang,

> > > If I open it on GNustep with TextEdit.app the german letters are 
> > > gone. If I
> > > create a file under GNUstep and open it on a Mac (TextEdit.app), the 
> > > german
> > > characters are replaced with garbage!?
> >
> > After looking at the attached rtf and doing some research, it looks 
> > like the encoding is "Mac OS Roman" (cpg1000):
> >
> >         http://en.wikipedia.org/wiki/Code_page_10000
> >
> > and not unicode.
>
> Right. And since Windows copepage 1252 is close to ISO 8859-1 (and 
> thus the
> first 256 characters of Unicode, the following simple patch will make 
> GNUstep
> write RTF files that -- at least as far as ISO 8859-1 characters are
> concerned -- are read correctly by OS X's TextEdit.app.
>
> --- TextConverters/RTF/RTFProducer.m.orig	2006-08-30 
> 17:59:51.000000000 +0200
> +++ TextConverters/RTF/RTFProducer.m	2006-08-30 18:00:29.000000000 
> +0200
> @@ -395,7 +395,7 @@
>    // grok paragraph spacing \saN. Should be no problem with other RTF 
> parsers
>    // as this command will be ignored. So this is for compatibility 
> with OS X.
>    result = (NSMutableString *)[NSMutableString stringWithString:
> -      @"{\\rtf1\\ansi\\ansicpg10000\\cocoartf102"];
> +      @"{\\rtf1\\ansi\\ansicpg1252\\cocoartf102"];
>
>    [result appendString: [self fontTable]];
>    [result appendString: [self colorTable]];

Works fabulous. Thanks a lot!

> The patch on the input side is more complicated, as it requires 
> reading the
> code page from the RTF file (which according to my cursory look is 
> currently
> ignored in the RTF parser) and setting up an appropriate iconv 
> translation in
> RTFConsumer.m (which is currently both beyond my time and iconv 
> expertise).

Fred Kiefer provided a hack for the other direction that is most likely 
improvable but works for now.

        pico ./TextConverters/RTF/rtfScanner.c

#include <iconv.h>

// read in a character as two hex digit
static int gethex(RTFscannerCtxt *lctxt)
{
  int c = 0;
  int i;

  for (i = 0; i < 2; i++)
    {
      int c1 = lexGetchar(lctxt);

      if (!isxdigit(c1))
        {
          lexUngetchar(lctxt, c1);
          break;
        }
      else
        {
          c = c * 16;
          if (isdigit(c1))
            {
              c += c1 - '0';
            }
          else if (isupper(c1))
            {
              c += c1 - 'A' + 10;
            }
          else
            {
              c += c1 - 'a' + 10;
            }
        }
    }

  {
    iconv_t cd;
    unsigned char c1 = c;
    unsigned char *i;
    unsigned int u;
    unsigned int *o;
    unsigned int inlen = 1;
    unsigned int outlen = 4;
    size_t rval;

    cd = iconv_open("UTF-16LE", "MACINTOSH");
    if (cd != (iconv_t)-1)
    {
        i = &c1;
        o = &u;
        rval = iconv(cd, (void*)&i, &inlen, (void*)&o, &outlen);
        if (rval != (size_t)-1)
        {
            c = u;
        }
        iconv_close(cd);
    }
  }

  return c;
}

Thanks guys! One more stone removed.

Regards,

  Andreas