On Sat, Jun 07, 2008 at 02:38:16PM +0200, Taco Hoekwater wrote: Hi all, I've forwaded communication on tex-hyphen to David. AFAIK -- when alphabet does not have more than 255 chars, opatgen's patterns are "isomorphic" to patgen's.
-- opatgen is necessary when one needs more than 255 chars, thus is not needed for German. -- you can have better / smaller patterns closer to the optimal ones by good setting of [o]patgen parameters (in my experience, four levels are enough, having higher numbers for "exceptions"). current 8 levels are definitely too much. have you tried setting from my papers for Czech/Slovak? I can help with that. All the best Petr > Stephan Hennig wrote: > >Taco Hoekwater schrieb: > > > >>I may be missing something here, but couldn't you use opatgen > >>(a patlib frontend) and do the whole process in unicode? > >> > >>See http://www.fi.muni.cz/~xantos/patlib/ > > > >Are patterns generated by patlib superior to patgen (in terms of overall > >size, number of needed levels, etc.) or is it merely patgen implemented > >as a library? > > Generated patterns should be very close (likely identical) to the > ones created by patgen, assuming the input abides the restrictions > imposed by patgen (it has to be 8bit and you have to make sure the > options you specify do not make the computations exceed patgen's > hardcoded internal limits). > > I say 'should' because I do not have intimate knowledge of either > program. David Antos will be the right person to ask for sure > (assuming you can get a hold of him, of course). > > Best wishes, Taco
