[XeTeX] Ligature unavailable in Estrangelo Antioch

[Jonathan Kew]

On 16 May 2009, at 19:25, Gareth Hughes wrote:

> Adam Twardoch wrote:
>>
>> The situation you're describing clearly indicate that the three
>> different ligatures should be placed in the Stylistic Alternates
>> ("salt") feature using the GSUB LookupType 3 (one-to-one-out-of- 
>> many).
>> In addition, you could create two Stylistic Sets ("ss01", "ss02")  
>> that
>> provide GSUB LookupType 1 (one-to-one) substitutions for the three
>> additional ligatures. Of course the lookups associated with those
>> features should be defined after the lookups that define the "liga"  
>> or
>> "rlig" features that turn on the ligatures.
>
> This is what I'd thought. The font has alternative ligature glyphs,  
> but
> they are not accessible as sequences of Unicode characters. It would  
> be
> far better, from the point of view of a XeTeX user, to have these
> alternatives accessible via salt or ssNN. However, I've never  
> understood
> how salt can cover one-to-many substitutions — how do I choose which
> one? At least with ssNN sets, I can choose a particular set. Can I
> choose multiple ssNN sets in XeTeX, and, if so, how are clashes dealt
> with — last called set overrides, or highest-numbered lookup table?

Sorry, you're not going to be able to do this with the current xetex  
OpenType implementation. For "complex" scripts (those such as Arabic,  
Syriac, or Indic scripts that have specialized "shaping engines"),  
there is not yet any support for specifying additional features; that  
is only supported for the "generic" engine that handles Latin and  
similar scripts.

As for one-of-many lookups, you'd do that by passing a parameter to  
the feature tag, as in

   \font\xyz = "My Font:salt=2"

to get the third alternate (they're indexed from zero). That should  
work in a Latin font with a salt feature, but it won't currently work  
in conjunction with Syriac.

And about multiple ssNN sets: each of the ssNN features is entirely  
separate, and you can enable as many of them as you like. The end  
result will depend on how the font developer has organized the lookups  
(and *not* on the order you list the features). It's up to the font  
designer to decide how "overlapping" features should interact if they  
are requested together.

JK