[luatex] Italic correction for normal text with LuaLaTex

[Reinhard Kotucha]

On 2012-03-12 at 01:19:12 +0100, Heiko Oberdiek wrote:

 > On Sun, Mar 11, 2012 at 09:47:06PM +0100, Reinhard Kotucha wrote:
 > 
 > > On 2012-03-11 at 20:54:32 +0100, Bernhard Tempel wrote:
 > > 
 > >  > Am 11.03.2012 18:48, schrieb Khaled Hosny:
 > 
 > >  > >  Try \setmainfont[RawFeature={+itlc}]{Latin Modern Roman}, it
 > >  > >  seems to make a noticeable improvement.
 > >  > 
 > >  > Thanks a lot for this. It really makes an improvement. In some
 > >  > cases the guessed glue is to big, but better results would be very
 > >  > difficult to reach when using heuristics. To calculate the extreme
 > >  > "\emph{f}!" combination works fine; with the less extreme
 > >  > combination "\emph{f}:" the kerning could be more tight, but I can
 > >  > see that the guessed value must be the the same as for "\emph{f}!".
 > > 
 > > But if you try
 > > 
 > >   \parindent0pt
 > >   \emph{f}! \par 
 > >   \emph{f}: \par 
 > >   \emph{f}; \par 
 > >   \emph{f}. \par  
 > > 
 > > you'll see that the period is treated differently.  Is this a result
 > > of the heuristics or does it come from somewhere else?
 > 
 > >From LaTeX. It manages a list \nocorrlist with the entries `,'
 > and `.' as default. \emph suppresses the italic correction
 > before a character in \nocorrlist.

Thanks for the hint.  It explains a lot.  Maybe for : and ; LaTeX
should provide a \halfcorrlist as well. :)

However, things are more difficult, of course, especially if you are
going to emphasize a part of a composed word.  You can't guess the
amount of italic correction needed from the bounding box.  Imagine the
characters d and b, which have the same height.  On the other hand it
seems that Khaled's heuristics are sufficient here:

 \emph{b}d \par
 \emph{d}b \par
 \emph{b}b \par
 \emph{d}d \par

The output is at least reasonable and going much further doesn't make
much sense if the information isn't provided by the fonts themselves.

Regards,
  Reinhard

-- 
----------------------------------------------------------------------------
Reinhard Kotucha                                      Phone: +49-511-3373112
Marschnerstr. 25
D-30167 Hannover                              mailto:reinhard.kotucha at web.de
----------------------------------------------------------------------------
Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
----------------------------------------------------------------------------