[tex-hyphen] More on hyphenating Ancient Greek.

[Jonathan Kew]

On 13/11/14 10:38, Philip Taylor wrote:
> Dear Claudio --
>
> Claudio Beccari wrote:
>
>> The problem is not LaTeX, but the program used for transforming XML
>> source into LaTeX code.
>
> There is no such program.  The XML is processed directly by XeTeX (see
> http://www.eutypon.gr/eutypon/pdf/e2013-31/e31-a02.pdf).
>
>> The additional characters and the scholarly emendations are dealt
>> with by LaTeX by means of package teubner.
>
> Neither LaTeX nor Tuebner are being used.
>
>> For what concerns Greek your problem probably persists even if you
>> use OpenType fonts, instead of the LGR encoded ones;
>
> I am not using LGR encoding; I am using Palatino Linotype "out of the
> box" (i.e., as installed in my Windows fonts directory with no adjunct
> files to re-map anything).
>
>> with the latter ones the round and angle brackets are mapped to other
>> chars and interfere with hyphenation. With OpenType fonts it is
>> possible that assigning a positive \lccode to round and angle
>> brackets hyphenation is still possible, but with unexpected results.
>
> Palatino Linotype is supplied in Truetype format :  if I were to play
> with the \lccodes for round and angle brackets, would I also have to
> re-generate the patterns ?
>

Yes. It's likely the existing patterns would find some hyphenation 
points in such text, but it's also likely some of them would be 
inappropriate, because the presence of these extra "letters" in the word 
could prevent some of the "inhibiting" patterns from matching.

(Xe)TeX doesn't have a way to simply ignore certain characters within a 
word for hyphenation purposes, and proceed to hyphenate the surrounding 
word as if they were not present. If they're "non-letters", they can't 
be part of a hyphenatable word (so only the separate fragments can be 
considered); and if they are "letters", you need patterns that were 
designed to take account of them as part of the spelling of the words to 
be hyphenated.

This isn't really a Greek issue, it's more general. For an English 
analogy, compare the results (in Plain TeX) of

   \showhyphens{colorful}
   % yields "col-or-ful"

   \showhyphens{colourful}
   % yields "colour-ful"; the en-US patterns don't do "col-our"

   \showhyphens{colo[u]rful}
   % doesn't find any hyphens; in particular, NOT "colo[u]r-ful"

   \lccode`[=`[  \lccode`]=`]
   \showhyphens{colo[u]rful}
   % yields "colo[u]r-ful", but other side-effects are a real risk,
   % so I can't recommend this as a general solution


I'm sure LuaTeX could be programmed to deal with this somehow.... :)

JK