[XeTeX] Accessing ligatures from FontForge

[Peter Baker]

On 2/23/11 9:27 AM, Meho R. wrote:
> That page provides an algorithm that is supposed to be used to convert
>> glyph names into sequences of Unicode character points.  If your glyphs
>> are named according to it, and if the PDF software follows it too, then
>> PDF software is supposed to be able to figure out what characters a glyph
>> represents for search purposes, even if it's not a standard ligature.  Of
>> course, there's no guarantee that a given package really will support the
>> rules properly.
>>
> Thanks for the link. However, even Adobe's OTF fonts have same problems
> when used with XeLaTeX regarding ligatures and searchability, so I don't
> think it is a naming convention issue. Curiously, when OTF fonts are
> used with Scribus and ligatures are inserted manually, they are
> recognized in PDF and no searchability issue there. Also, when OTF fonts
> are converted to TTF, it seems searchability issue is gone with XeLaTeX
> too (at least seems like that for couple of fonts I just tested).

This is an old problem. Whether or not search works properly in the most 
widely used PDF readers depends on the font. As noted here, the ligature 
glyphs have got to be properly named so that the PDF software can derive 
the names of the component glyphs. But *in addition* it is necessary 
that the ligature glyphs be

1. unencoded; or
2. one of the few ligatures with Unicode encodings (fi, fl, etc.).

If non-standard ligatures have been assigned code-points in the font's 
Private Use Area, most PDF software won't try to analyze the glyph name. 
The thinking seems to be: if the point of such analysis is to derive one 
or more Unicode encodings, what's the point if the glyph already has one?

At least one person has mentioned to me that some PDF software does 
better, but I haven't observed this myself. And workarounds have been 
discussed on this list.

Peter