Hi!

David Faure and I are developing a replacement of the current MS Word filter
in KOffice. The project's name is wv2 (the 2nd version of the word-view
library) and consists of two parts: a document reader library, which handles
the gory parsing of the MS Word documents and feeds a so-called consumer
with the gathered information. This abstraction makes it possible to use the
low-level library in other projects, too.

wv2, the low-level library, depends on a working C++ compiler (tested with
gcc-2.95.3, the Mandrake gcc-2.96, gcc 3.1, and gcc-3.2) and libiconv
(should be installed on almost all systems).
Additionally it uses libole2, but if this library is not found on your
system it will use an internal copy (including a stripped down version of
glib). Shortly put -- very light dependencies, so it shouldn't be a major
headache to install wv2.

The library is also designed to run on various systems, but I only have
access to the sourceforge compile farm, so please give it a try with your
weird compilers and Unices (and report any failures). It should be pretty
portable, but the compiler has to support namespaces and provide a working
STL.

Currently the filter is able to import section, paragraph and character
properties, tabs, lists, and several other properties from Word 97 and
Word 2000 documents (I didn't test Word XP, but it should also work). I
put a screenshot showing some of those features here 
http://www.sbox.tugraz.at/home/t/trobin/wv2_1.png and here
http://www.sbox.tugraz.at/home/t/trobin/wv2_2.png.
It's planned to support Word 6 and Word 95 in the near future. I also have
some documentation about older formats (Word 2, 3, 4, 5), but I consider
support for other features like tables or embedded images more important.

We think the filter has reached a stage where curious developers and users
might want to give it a try. This would help us to test the filter on real
world documents, not just on our limited set of test documents.


How to install it:
1) Get the "wv2" module from CVS http://sourceforge.net/cvs/?group_id=10501
   or download the tarball at
   http://www.sbox.tugraz.at/home/t/trobin/wv2-20020910.tar.bz2 (471KB)
2) Install it (./autogen.sh && make && make install)
   Please don't use any special --prefix for now, as the configure check in
   KOffice doesn't allow to specify alternative locations, yet.
3) Update koffice/filters/kword/msword and rerun make -f Makefile.cvs in the
   toplevel directory.
4) Run configure, it should list a (successful) check for wv2. To verify that
   you could grep your config.h for HAVE_WV2 (should be #defined)
5) make / make install the koffice/filters/kword/msword directory
6) Remove $KDEDIR/share/services/ole_winword97_import.desktop (this step is
   very important, if you forget about that you'll just use the old filter)
   Alternatively you could also update koffice/filters/olefilters and just
   execute make install again. This will uninstall the .desktop file for you.
7) Run kbuildsycoca or trust the dirwatcher thingy to trigger it for you.

Use it like any other KOffice filter.

What to test:
1) Please only test Word97 (or newer) documents. Opening any older documents
   will return an empty KWord document for now.
2) Watch out whether it imports all the text with the correct properties
   (like the font, bold/italic/underlined, lists (to the extend KWord supports
   them), tabs,...).
3) Please report all crashes you get, we are currently not able to crash the
   filter with any of our test documents.

In case you report bugs or crashes, please include a full log (and, if
possible, also attach the Word document causing the crashes) and send the bug
report to Werner Trobin <trobin@kde.org>


Please do not report that tables, images, or embedded documents are not
correctly imported. We know that already :-)

I suggest that all the installation troubleshooting should be done on the
koffice@kde.org list, as it's very likely that many testers have the same
problem.

Thanks for your attention,
David Faure
Werner Trobin

P.S. The filter is a bit on the slow side right now, as we're using QDom*
in the consumer filter (about two thirds of the time are wasted in code
related to QDom construction/serialisation). This will change soon.

____________________________________
koffice mailing list
koffice@mail.kde.org
To unsubscribe please visit:
http://mail.kde.org/mailman/listinfo/koffice