Work plan for hyphenation in OpenOffice

• See also this Howto file from OpenOffice

An OpenOffice.org hyphenator was not a fixed part of the original plan, rather it was indicated as an optional addition. Part of the reason for this was that at the time I could not find information on how to implement it. Now the information is there (or: we have found it), and we’ll try to add Sámi hyphenation after we have the OOo speller in place. Relevant links are:

• OOo Lingucomponent, hyphenation, includes a brief how-to
• OOo Dictionary Wiki, with link to hyphenation dictionary creation tool
• A copy of the user manual for OPatGen (see below) is available.

Basically, the task of creating TeX hyphenation pattern files consis of the following steps:

1. print the lower side of a hyphenated, normative transducer to produce a correctly hyphenated word list
2. run that list through OPatGen (see above) and associated scripts - see the user documentation above
3. test the result

We need to install OPatGen first, which in turn needs cweave.

Steps to install cweave:

sudo port install texlive-bin-extra

Steps to install opatgen (we have a local copy of the code):

1. cd $GTHOME/tools/patlib

2. make
   NB! At the moment the C compilation is broken, because the code rely on a number of old, non-conformant coding practices. gcc 2.95 is required, which is not available on any modern systems anymore. We either need to upgrade the code, or find an alternative solution.

   We have successfully modernised the code of tools/dic2traskelet.C (a helper tool), but the C code of the main application is a bit more complex to cover.

The alternative solution

The OPatGen route is too time-consuming, so we need an alternative. Probably the simplest alternative is to use the original PatGen tool, which differs from OPatGen in two important respects:

• it does not support Unicode, so all non-ASCII chars must be escaped
• it comes with almost every TeX installation, and should be available everywhere

The steps needed to create the hyphenation patterns are almost the same, with the addition of the need to escape the non-ASCII characters. In more detail, it looks like the following: