[xmlroff] text-align: justify

[Giacomo C. <jw...@bu...>]

Dear xmlroff list,
I'm posting here a brief summary of what I found out recently regarding
justification of paragraphs (sorry Tony, I didn't notice your request at
first, due to scrolling-hurry):

1) Since Pango didn't support justification natively, Damon Chaplin has
written a patch which adds justification, however the initial patch had
some problems, so he sent me a new one which should be almost final (in
attachment). The full justification story can be found here:
http://bugzilla.gnome.org/show_bug.cgi?id=64538

2) After applying the above patch, xmlroff could be modified to support
justification, so I wrote a small patch (in attachment). Please notice
that text-alignment and direction is still in early stage and most
combinations aren't supported. I suppose xmlroff needs some kind of
masterplan to face all the possible orientations/alignments. I'm willing
to contribute, however I need some time and training on pango etc. ;-)

3) The result isn't completely satisfactory due to the following problem:

(pseudo-quoting from a private mail to Tony Graham)

---> snip <---

When a 3-line paragraph is compressed into a 2-line paragraph, the
following assertion shows up:

process:26210): libfo-CRITICAL **: fo_doc_gp_do_callbacks: assertion
`line_last >= line_first && line_last <= g_slist_length
(pango_layout_get_lines (layout)) - 1' failed

(process:26210): libfo-CRITICAL **: fo_doc_gp_do_callbacks: assertion
`line_last >= line_first && line_last <= g_slist_length
(pango_layout_get_lines (layout)) - 1' failed

and an empty line is
showed after the paragraph. I suppose for some reason the real number of
lines isn't updated correctly. Damon tells me that he doesn't experience
the above problem using pango with GtkTextView widget, so could you
please verify how/when lines count are fetched from pango and what
actually happens if pango returns a shorter line count?

---> snip <---

Tony answers that:

---> snip <---

The assertion happens in the xmlroff code in libfo/fo-doc-gp.c:

------------------------------------------------------------
static void
fo_doc_gp_do_callbacks (GnomePrintContext *context,
            PangoLayout       *layout,
            gint              line_first,
            gint              line_last,
            gint              x,
            gint              y)
{
  PangoLayoutIter *iter;
 
  g_return_if_fail (context != NULL);
  g_return_if_fail (PANGO_IS_LAYOUT (layout));
  g_return_if_fail (line_first >= 0);
  g_return_if_fail (line_last >= line_first &&
            line_last <= g_slist_length (pango_layout_get_lines
(layout)) - 1);

  iter = pango_layout_get_iter (layout);

  gint line_number = -1;
  do
    {
      PangoRectangle   logical_rect;
      PangoLayoutLine *line;
      int              baseline;
     
      line_number++;

      if (line_number < line_first)
    {
      continue;
    }

      line = pango_layout_iter_get_line (iter);
     
      pango_layout_iter_get_line_extents (iter, NULL, &logical_rect);
      baseline = pango_layout_iter_get_baseline (iter);
     
      fo_doc_gp_do_line_callbacks (context,
                   line,
                   x + logical_rect.x,
                   y - baseline);

      if (line_number >= line_last)
    {
      break;
    }
    }
  while (pango_layout_iter_next_line (iter));

  pango_layout_iter_free (iter);
}
------------------------------------------------------------

The assertion is to stop you trying to lay out more lines than are in the
PangoLayout.

xmlroff handles printing subsets of the lines in a PangoLayout since a
PangoLayout may be broken across a page.

I suspect that xmlroff is getting a count of the lines before the
justification code removes the last line.

I would expect that the do-while in the function only iterates for as many
lines as there are in the PangoLayout and that what you're seeing as an
extra
line is because xmlroff is leaving enough room for the original 3-line block
and not the current 2-line block.

---> snip <---

I suppose that's about all.

Sincerely,

- Giacomo Cariello