[XeTeX] Fwd: Text figures (Old style figure) in XeTeX

Jonathan Kew jonathan_kew at sil.org
Wed Dec 19 20:28:18 CET 2007

On 19 Dec 2007, at 5:58 pm, kmc.best wrote:

> ---------- Forwarded message ----------
> From: "kmc.best" <kmc.b... at gmail.com>
> Date: Dec 19, 6:54 pm
> Subject: Text figures (Old style figure) in XeTeX
> To: comp.text.tex
> Hi all,
> I'm using XeTeX 0.997 on MiKTeX 2.7 to explore some features of a
> professional font (Minion Pro):
> \usepackage[cm-default]{fontspec}
> \setmainfont[Numbers=OldStyle]{Minion Pro}
> I found that the text figures can be displayed in the resulting PDF
> but they can't be searched/copied, which indicates that they're not
> well embedded. Is this a know issue?

I think this is happening because Adobe encodes these glyphs in the  
Private Use Area of Unicode, rather than leaving them unencoded and  
relying solely on OpenType features to access them. Because of this,  
the ToUnicode mapping that xdvipdfmx embeds in the PDF will map these  
to PUA codepoints (U+F643..F64C, in the case of the OldStyle  
numerals) rather than the proper digit codepoints.

IMO, this is a poor design choice by the font developer; they should  
not be using PUA character codes for things that are not distinct  
characters but glyph variants of existing standard characters. To  
some extent, it may be a legacy of the pre-OpenType days when every  
glyph had to be directly encoded in some way, in order to be  
accessible (hence "expert sets", etc.). In these days of Unicode and  
OpenType, this is no longer necessary or appropriate.

It may be possible to modify xdvipdfmx's algorithms for ToUnicode  
generation to handle such fonts better; I'll look into it when time  


More information about the XeTeX mailing list