Re: [Groff] Re: new grotty format

[Akim Demaille]

>>>>> "Rick" == Rick Richardson <address@hidden> writes:

Rick> OTOH, at least some of those programs which read Groff's output
Rick> don't handle the defacto overstrike conventions, either.  For
Rick> example, "enscript" recognizes underline but not bold. "a2ps"
Rick> recognizes bold but not underline.

Heck, I must have missed that sequence.  What form of `underline' are
you referring to?  a2ps recognizes this:

/*-------------------------------------------------------.
| Returns a single char (EOF for end-of-file).           |
|                                                        |
| Does the nroff's replacement for underline, bold etc.  |
`-------------------------------------------------------*/

static int
plain_getc (buffer_t * buffer, enum face_e *face)
{
  uchar c;

  if (buffer_is_empty (buffer))
    {
      buffer_get (buffer);

      /* We don't trust liba2ps for the line numbers, because if a2ps
         skips some lines (e.g., --strip-level, or INVISIBLE), liba2ps
         will number upon output lines, not imput lines, which is what
         is expected.  */
      (CURRENT_FILE (job))->lines = buffer->line;

      if (buffer->len == 0)
        return EOF;
    }

  *face = Plain;
  c = buffer->content[(buffer->curr)++];

  /* Check if it is a special nroff'ed sequence */
  if (buffer->content[buffer->curr] == '\b')
    {
      /* We might be dealing with misceleanous nroff'ed pages. */
      const uchar *input = buffer->content + buffer->curr + 1;

      /* This might be a bolding sequence.  The bad news is that some
         strange systems build the bold sequences with only one
         rewriting, not the 3 usuals.

         Super strong `_', seen in Sun's mpeg_rc doc.  */
      if (c     == input[0] &&
          '\b'  == input[1] &&
          c     == input[2] &&
          '\b'  == input[3] &&
          c     == input[4] &&
          '\b'  == input[5] &&
          c     == input[6])
        {
          *face = Label_strong;
          buffer->curr += 8;
        }
      else if (c     == input[0] &&
               '\b'  == input[1] &&
               c     == input[2] &&
               '\b'  == input[3] &&
               c     == input[4])
        {
          *face = Keyword_strong;
          buffer->curr += 6;
        }
      else if  (c    == input[0] &&
                '\b' == input[1] &&
                c    == input[2])
        {
          *face = Keyword_strong;
          buffer->curr += 4;
        }
      else if  (c == input[0])
        {
          *face = Keyword_strong;
          buffer->curr += 2;
        }
      /* If C is `_', then set font to italic and move to next char.
         */
      else if (c == '_')
        {
          char c2 = input[0];
          /* Winner of the cup: mpeg_rc, from Sun, where it tries both
             to underline, and to boldize.  */
          if    ('\b' == input[1] &&
                 c2   == input[2] &&
                 '\b' == input[3] &&
                 c2   == input[4] &&
                 '\b' == input[5] &&
                 c2   == input[6])
            {
              *face = Label_strong;
              c = c2;
              buffer->curr += 8;
            }
          else
            {
              *face = Keyword;
              c = input[0];
              buffer->curr += 2;
            }
        }
      /* Seen in gcc.1: o;\b;+, seen in cccp.1: +;\b;o to have an
         itemizing symbol.  */
      else if (('o' == c &&
                '+' == input[0])
               || ('+' == c &&
                   'o' == input[0]))
        {
          *face = Symbol;
          buffer->curr += 2;
          c = 0305; /* \oplus in LaTeX */
        }
      /* Seen in groff.1 : c;\b;O, for copyright */
      else if ('c' == c &&
               'O' == input[0])
        {
          *face = Symbol;
          buffer->curr += 2;
          c = 0343; /* \copyright. */
        }
      /* Seen in gtroff.1 : +;\b;_, for plus or minus */
      else if ('+' == c &&
               '_' == input[0])
        {
          *face = Symbol;
          buffer->curr += 2;
          c = 0261;
        }
      /* Seen in geqn.1 : ~;\b>;\b;_ for greater or equal */
      else if ('~'  == c &&
               '>'  == input[0] &&
               '\b' == input[1] &&
               '_'  == input[2])
        {
          *face = Symbol;
          buffer->curr += 4;
          c = 0263;
        }
      /* Less than or equal to. */
      else if ('~'  == c &&
               '<'  == input[0] &&
               '\b' == input[1] &&
               '_'  == input[2])
        {
          *face = Symbol;
          buffer->curr += 4;
          c = 0243;
        }
      /* Underlined: x;\b;_ . Note that we have a conflict here in the
         case x == '+' (see above).  This choice seems the best.  */
      else if ('_' == input[0])
        {
          *face = Keyword;
          buffer->curr += 2;
        }
      /* (Should be last). In some case, headers or footers too big,
         nroff backslashes so that both chars. are superimposed.  We
         decided to keep only the first one.  */
      else if (((CURRENT_FILE (job)->lines + 3) % 66 == 0)
               || ((CURRENT_FILE (job)->lines - 4) % 66 == 0))
        {
          buffer->curr += 2;
        }
      /* else: treate the backslash as a special characters */
    }
  return c;
}